Accruals and the performance of stock returns following external financing activities *

Transcription

1 Accruals and he performance of sock reurns following exernal financing aciviies * Georgios Papanasasopoulos Deparmen of Banking and Financial Managemen of he Universiy of Piraeus Deparmen of Economics of he Universiy of Ahens papanas@unipi.gr Dimirios Thomakos Deparmen of Economics of he Universiy of Peloponnese homakos@uop.gr Tao Wang Deparmen of Economics of he Ciy Universiy of New York ao.wang@qc.cuny.edu Firs Draf: Augus 15, 008 This Draf: July 13, 009 * The auhors hank Gikas Hardouvelis and Emmanuel Tsiriakis for insighful commens and suggesions. The usual disclaimer applies. Corresponding Auhor

2 Accruals and he performance of sock reurns following exernal financing aciviies Absrac: In his paper we find ha he close relaion of he anomalies on exernal financing aciviies wih he accrual anomaly is mainly aribuable o invesing capial accruals. However, he predicive power of exernal financing aciviies for fuure sock reurns is found unrelaed o ha of working capial accruals. Working capial accruals play an imporan role only on he predicabiliy of sock reurns, following shor erm deb financing aciviies. Overall, our evidence is more likely o be consisen wih invesor s failure o recognise opporunisic earnings managemen and/or agency relaed overinvesmen associaed wih invesed capial. Keywords: Exernal financing aciviies, accruals, sock reurns JEL classificaion: G10, M4

3 1. Inroducion An exensive body of research documens a negaive relaion beween ne exernal financing aciviies and fuure sock reurns. Aciviies raising new capial are associaed wih lower fuure sock reurns, while aciviies disribuing capial are associaed wih higher fuure sock reurns (Rier 1991, Loughran and Rier 1995, Loughran and Rier 1997, Spiess and Affleck-Graves 1999, Bille e al. 001, Ikenberry e al. 1995, Michaely e al. 1995, Affleck-Graves and Miller 006, Daniel and Timan 006, Poniff and Woodgae 008, Fama and French 008). However, he economic raionale of he subsequen drif in reurns remains a conroversial issue. Under a behavioral inerpreaion, Rangan (1998), Teoh e al. (1998) and Heron and Lie (004) offer he hypohesis ha managers are engaged in opporunisic earnings managemen around equiy offerings by exploiing (discreionary) accruals, in order o increase he offering proceeds. Invesors fail o recognize earnings managemen and naively exrapolae ransiory earnings increases, resuling in an overvaluaion of issuing firms. 1 Under a raional inerpreaion, Shivakumar (000) offers he managerial response hypohesis: earnings managemen hrough (discreionary) accruals by equiy issuers reflecs a raional response o anicipaed marke behavior a offering announcemens. Since issuers canno credibly signal he absence of earnings managemen, invesors rea hem uniformly as having inflaed prior earnings and raional discoun heir sock prices. A more recen, bu growing, lieraure focus on he exernal financing effec by using measures from he saemen of cash flows. In paricular, Bradshaw e al. (006) have consruced a parsimonious measure of he ne amoun of cash generaed by corporae financing (equiy and deb) aciviies and show ha his measure is negaively relaed wih fuure sock reurns. They also show a negaive (posiive) relaion beween ne exernal financing and fuure earnings performance (over-opimism in analyss forecass). As saed by Bradshaw e al. (006), marke iming and opporunisic earnings managemen are wo compeing explanaions for heir findings. As a hird explanaion, managers could inves he proceeds from heir exernal financing aciviies in value-desroying projecs o serve heir own ineress (agency relaed overinvesmen). When invesors learn ou ha such expendiures dissipae firm value, sock prices adjus downward. Neverheless, he above evidence on he exernal financing anomaly could be closely relaed o he accrual anomaly. The accrual anomaly, firs documened by Sloan (1996) refers o he empirical regulariy ha low-accruals firms experience higher fuure sock reurns han 1 An alernaive behavioral inerpreaion is based on he marke iming hypohesis of Loughran and Rier (1995): Firms end o issue (repurchase) securiies when hey are overvalued (undervalued). An alernaive raional inerpreaion is offered by Eckbo e al. (000): equiy issuers have lower defaul risk, and hus are priced o yield lower expeced reurn. 3

4 high-accruals firms. Accouning accruals represen he difference beween a firms accouning earnings and is underlying cash flows. Accruals improve earnings abiliy o reflec firm performance in ha hey minimize iming and maching problems inheren o cash flows. 3 Accruals also allow for imely recogniion of gain and losses due o unanicipaed revisions of expeced fuure cash flows 4, albei in an asymmerical fashion. 5 Neverheless, as observed by boh heoreical and pracical exs he beneficial role of accruals is reduced o he exen ha managers manipulae earnings hrough accruals. In his line, issuers ha are engaged in earnings manipulaion are more likely o have high accruals and earn low reurns. Furher, accouning accruals represen growh in ne operaing asses on a firms balance shee. Firms balance shee consrain implies ha he sources of funds mus be equal o he uses of funds. Therefore, even if managers do no manipulae earnings, firms raising capial are likely o have high accruals and earn low fuure sock reurns, and firms disribuing capial are likely o have low accruals and earn high fuure sock reurns. As a resul, accruals may have an imporan role in inerpreing exernal financing anomaly. This issue has been firs ackled by Cohen and Lys (006) who show ha afer conrolling for oal accruals, he negaive relaion beween exernal financing aciviies and fuure sock reurns is aenuaed and no saisical significan. As argued by Cohen and Lys (006), heir findings are more consisen wih agency relaed over-invesmen raher han marke iming by managers. In follow up research, Dechow e al. (008) offered a similar inerpreaion for he exernal financing anomaly. 6 I is obvious from he above findings ha accruals could be a key in undersanding he exernal financing anomaly. Based on Richardson e al. (005) accruals can be divided o working capial accruals and invesing capial accruals. Working capial accruals are relaed o operaing aciviies and may arise from discreion over accouning rules wih respec o he naure, iming and magniude of revenues and expenses recogniion (e.g. premaure booking of sales, allocaion of more overhead expenses o invenory han o cos of goods sold). Thus, one canno rule ou he possibiliy ha managers can inflae earnings by recording (discreionary) working capial accruals during periods in which hey raise exernal financing. In his line, if earnings managemen is driven by opporunism or hubris o mislead invesors, he exernal financing anomaly could arise as he sock marke emporarily overvalues issuing firms and is subsequenly dissappoined by unexpeced declines in earnings. On he oher hand, if earnings managemen is a response o anicipaed marke behavior, he exernal financing anomaly could arise as invesors lower heir assessmens of prior earnings surprises of issuing firms and raionally discoun heir sock prices. 3 See Dechow (1994), Guay, Kohari and Was (1996), Dechow, Kohari and Was (1998). 4 See Basu (1997), Ball, Kohari and Robin (000), Ball, Robin and Wu (000, 003). 5 The asymmery arises from he conservaive naure of GAAP, where losses are recognized immediaely and he recogniion of gains is deferred o he fuure unil realized. 6 A similar explanaion has been also offered by Richardson and Sloan (003). 4

5 Invesing capial accruals are relaed wih invesmen aciviies and could be derived from discreion over accouning rules (e.g. capializaion of operaing expenses as fixed asses, subjecive esimaion of long erm receivables). Thus, here is a possibiliy of earnings managemen by execuives of issuing firms hrough (discreionary) invesing capial accruals. In his line, he exernal financing anomaly could be again consisen wih invesor s failure o recognise opporunisic earnings managemen or invesor s raional correcion for lower earnings qualiy. A he same ime, firms could have invesing capial accruals if heir managers (wih empire building incenives) engage in waseful spending by using he ne proceeds from exernal financing aciviies. Thus, he exernal financing anomaly could be a consequence of marke undereacion o he informaion conained in possible overinvesmen. On he oher hand, Anderson and Garcia-Feijoo (006) argue ha capial invesmen may capure risk in growh opions. Hence, he exernal financing anomaly could be also compensaion for higher risk associaed wih low capial invesmen. From he above analysis, i is clear ha economic raionale of he negaive drif in sock reurns following ne exernal financing aciviies could differ, wheher hese aciviies are associaed wih working or invesing capial accruals. Previous research, has no generally disinguished beween he implicaions of accruals from disinc business aciviies on he exernal financing anomaly. The above issue moivaes wha we do in his paper. Our work is organized along wo dimensions. Firs, we invesigae he properies of porfolios and hedge sraegies based on he magniude of he ne amoun of cash generaed by enire and individual exernal financing aciviies. This le us assess he economic significance of he informaion in exernal financing aciviies for he cross secional variaion in sock reurns. Second, we invesigae wheher he anomalies on exernal financing aciviies capure he same underlying paern in sock reurns wih he anomalies on oal accruals, working capial accruals and invesing capial accruals. In his way, we can offer o he exising lieraure a deeper undersanding on he role of accruals on he predicabiliy of sock reurns following exernal financing aciviies. The remainder of he paper is organized as follows: Secion provides a deailed descripion of our research design. In secion 3 we presen daa, sample formaion, variables measuremen, while in secion 4 we provide our empirical resuls. Secion 5 summarizes and concludes he paper.. Research Design In his paper, we invesigae he role of accruals in inerpreing he subsequen drif in sock reurns following exernal financing aciviies. Following Bradshaw e al. (006), we use he parsimonious measure of he ne amoun of cash generaed by corporae financing aciviies ( XFIN, hereafer) ha allows us o focus on individual and enire corporae 5

6 financing ransacions. This measure is defined as he difference beween cash flows received from issuance of new equiy and deb financing (sock issues plus deb issues) and cash flows used for he reiremen of exising equiy and deb financing (sock repurchases plus dividend paymens minus deb repaymens). We hen, decompose i across balance shee caegories based on he naure of he underlying securiies ha are being issued and reired. In paricular, XFIN will be decomposed ino ne cash flows generaed from equiy financing aciviies ( EQUITY, hereafer) and deb financing aciviies ( DEBT, hereafer). XFIN = EQUITY + DEBT (1) EQUITY is defined as he difference beween cash flows received from sock issues and cash flows disribued for sock repurchases and dividends paymens. DEBT is defined as he difference beween cash flows received from deb issues and cash flows disribued for deb repaymens. However, we also disinguish beween ne shor and long erm deb financing aciviies since heir predicive power for fuure sock reurns could differ. Previous work has no generally disinguished beween differen forms of deb financing aciviies and heir effecs on sock prices. In paricular, DEBT will be also decomposed ino ne cash flows generaed from shor erm deb financing aciviies ( SDEBT, hereafer) and long erm deb financing aciviies ( LDEBT, hereafer) DEBT = SDEBT + LDEBT () SDEBT ( LDEBT) is defined as he difference beween cash flows received from shor (long) erm deb issues and cash flows disribued for shor (long) erm deb repaymens. To our knowledge, his is he firs paper in he lieraure ha focuses on he relaion beween shor and long erm deb financing aciviies wih fuure sock reurns. Our work is organized along wo dimensions. Firs, we invesigae financial and reurn characerisics of porfolios and hedge sraegies based on he magniude of he ne amoun of cash generaed by enire and individual exernal financing ransacions. Of course, we recognize ha one canno ignore risk in examining sock reurns. For his purpose, we follow oher sudies in he accouning lieraure on he exernal financing anomaly (Bradshaw e al. (006), Cohen and Lys (006) and Dechow e al. (008)) and consider in our analysis sizeadjused reurns. Noe, ha we also invesigae he robusness of our sock reurns ess by applying he saisical arbirage es designed by Hogan e al. (004) o hedge sraegies on all exernal financing measures. This es circumvens he bad model problem of sock reurn ess in he anomalies lieraure since is definiion is no coningen upon a specific model of marke reurns. In paricular, we es wo implicaions of saisical arbirage for each sraegy: wheher is mean annual incremenal profi is posiive and wheher is ime-averaged variance decreases over ime. To our knowledge, his is he firs paper ha examines wheher sraegies 6

7 on exernal financing measures consiue saisical arbirage opporuniies. Our work on he properies of exernal financing aciviies le us assess he economic significance of heir informaion for he cross secional variaion in sock reurns Second, we invesigae wheher he anomalies on exernal financing aciviies capure he same underlying paern in sock reurns wih he anomalies on oal accruals, working capial accruals and invesing capial accruals. The anomaly on working capial accruals has been firs documened by Sloan (1996), while on invesing capial accruals and oal accruals by Richardson e al. (005). Working capial (curren operaing) accruals CACC are defined as growh in ne working capial (ne curren operaing asses), invesing (non curren operaing) accruals NCACC as growh in ne invesed capial (ne non curren operaing asses) and oal accruals TACC as growh in ne operaing asses: TACC = CACC + NCACC (3) In our analysis on he relaion of he anomalies on exernal financing aciviies wih he anomalies on accruals, we consider conrol hedge and join hedge porfolio sraegies. Then, following Fama-MacBeh (1973), we also esimae separae cross secional reurn regressions on exernal financing and accrual measures, afer conrolling size and book o marke raio Daa, Sample Formaion and Variable Measuremen. Our sample covers all firms wih available daa on Compusa and CRSP for he period Moreover, we exclude all firm year observaions wih SIC codes in he range (financial companies) because he discriminaion beween operaing and financing aciviies is no clear for hese firms. Furhermore, we require as in Vuoleenaho (00) all firms o have a December fiscal year end, in order o align accouning variables across firms and obain radable invesmen sraegies for our subsequen porfolio assignmens. Finally, we eliminae firm year observaions wih insufficien daa on Compusa o compue he primary financial saemen variables used in our ess. 8 These crieria yield final sample sizes of 105,119 firm year observaions wih non missing financial saemen and sock reurn daa. Following Dechow e al. (008), we use he indirec mehod (balance shee) 7 As suggesed by Fama and French (1993) among oher, firm characerisics such as size and book o marke raio may help explain he cross-secional variaion of sock reurns since hey could proxy for ime-varying sysemaic risk. 8 In paricular, we eliminae firm year observaions if Compusa daa iems 1, 4, 5, 6, 18 and 181 are missing. If daa iems 9, 34, are missing, we se hem equal o zero raher han eliminaing he observaion. The resuls are qualiaively similar if we insead eliminae hese observaions. 7

8 mehod o measure exernal financing and accrual measures as follows 9 : EQUITY where: = ( TA TL ) NI NI : Ne income (daa iem 18). TA : Toal asses (daa iem 6). TL : Toal liabiliies (daa iem 181). ( ) SDEBT = STD where: STD : Shor erm deb (daa iem 34). ( ) LDEBT = LTD where: LTD : Long erm deb (daa iem 9). DEBT = SDEBT + LDEBT XFIN CACC where: = EQUITY + DEBT = ( CA C ) ( CL STD ) CA : Curren asses (daa iem 4). C : Cash and cash equivalens (daa iem 1). CL : Curren liabiliies (daa iem 5). NCACC = ( TA CA ) ( TL CL LTD ) TACC = CACC + NCACC Similar o prior sudies, XFIN, EQUITY, DEBT, TACC, CACC and NCACC are deflaed by conemporaneous average oal asses and hen winsorized a +1 and 1 in order o eliminae he influence of ouliers. As menioned in he previous secion, in our analysis, we also consider marke capializaion (MV) and book o marke raio (BV/MV) Marke capializaion is measured as price per share (iem 199) imes shares ousanding (iem 5) a he beginning of he porfolio formaion monh. Noe ha we require a leas a four-monh gap beween he porfolio formaion monh and he fiscal year end o ensure ha invesors have 9 As documened by Dechow e al. (008) he balance shee mehod requires clean surplus assumpions and ha all ineres expense is paid in cash. To check for robusness, we replicae all our empirical ess by using measures of corporae financing aciviies exraced from he cash flows saemen and find qualiaively similar resuls. However, daa from he cash flow saemen limi our sample size since hey are available from

9 financial saemen daa prior o forming porfolios. 10 Book o marke raio is defined as he raio of he fiscal year end book value of equiy (iem 60) o he marke capializaion. The annual one-year ahead raw sock reurns RET are measured using compounded 1- monh buy-hold reurns inclusive of dividends and oher disribuions from he CRSP monhly files. Then, size-adjused reurns SRET are calculaed by deducing he value weighed average reurn for all firms in he same size-mached decile, where size is measured as he marke capializaion a he beginning of he reurn cumulaion period. The size porfolios are formed by CRSP and are based on size deciles of NYSE and AMEX firms. If a firm is delised during our fuure reurn window, hen he CRSP s delising reurn is considered for he calculaion of he one-year ahead raw sock reurn, and any remaining proceeds are reinvesed in he CRSP value-weighed marke index. This miigaes concerns wih poenial survivorship biases. If a firm is delised during our fuure reurn window as a resul of poor performance and he delising reurn is coded as missing by CRSP, hen a delising reurn of - 100% is assumed. 4. Resuls 4.1 Characerisics from Exernal Financing Porfolios Table 1 repors ime series averages of annual mean values of exernal financing and accrual characerisics of porfolios formed on he magniude of ne exernal financing aciviies and heir componens. For his purpose, each year we rank firms independenly on ne exernal financing aciviies and heir componens, allocae hem ino en equal-sized porfolios (deciles) based on hese ranks and hen compue heir exernal financing and accrual characerisics. The porfolios are held for one year and hen rebalanced. Noe ha we require a leas a four-monh gap beween he porfolio formaion monh and he fiscal year end o ensure ha invesors have financial saemen daa prior o forming porfolios. We also repor he ime series averages of spreads in characerisics across he lowes and he highes decile, along wih he associaed -saisic. A he lowes decile firms are disribuing capial, while a he highes decile firms are raising capial. In Panel A of Table 1, we provide characerisics of porfolios formed on he magniude of ne exernal financing aciviies. From he firs row we see ha he ime series averages of XFIN for ne repurchasers and ne issuers are and , respecively, while he spread is (=.79). Turning, o he second and he hird row, we see ha ne repurchasers are more likely o repay deb, while ne issuers are more likely o issue deb. The spread of EQUITY and DEBT beween ne 10 Alford e al. (1994) argue ha four monhs afer he fiscal year end, all firm s financial saemen daa are publicly available. 9

10 repurchasers and ne issuers is 0.6 (=11.38) and (=4.33), respecively. This finding, consisen wih pecking order heory, indicaes ha firms are more likely o engage in deb han equiy financing aciviies. Resuls on he fourh and he fifh row reveal ha firms are more likely o engage in long erm han shor erm deb financing aciviies. The spread of SDEBT and LDEBT is (=3.56) and 0.17 (=31.36), respecively. Furhermore, we see an increasing rend in oal accruals, working capial accruals and invesing capial accruals across deciles. The ime series averages of TACC for ne repurchasers and ne issuers are and 0.308, respecively, while he spread is (=-41.15). From he las wo rows, we see ha ne repurchasers have similar in absolue value ime series averages of CACC and NCACC. However, ne issuers have higher in absolue value ime series averages of NCACC han CACC. In paricular, he ime series averages of CACC and NCACC for ne repurchasers are and , respecively, while for ne issuers are and 0.7, respecively. The spread of CACC and NCACC is (=-8.69) and (=-7.36), respecively. Overall, resuls from he las wo rows reveal ha firms wih high values of ne exernal financing are more likely o have high invesing capial accruals raher han working capial accruals. Panel B of Table 1, repors characerisics of porfolios formed on he magniude of ne equiy financing aciviies. The ime series average of EQUITY for equiy repurchasers and dividend paying firms is 0.13, for equiy issuers -0.43, while he spread is (=13.94). Noe also, ha equiy repurchasers and dividend paying firms are deb issuers, a finding indicaive of possible refinancing aciviy whereby he proceeds from deb issues could be used o repurchase equiy and pay dividends. Time series averages of SDEBT and LDEBT for hose firms are and respecively, suggesing ha firms are more likely o issue long erm han shor erm deb. Turning o equiy issuers, we see ha ha he ime series average of LDEBT is , while for SDEBT is no saisically significan. This finding, suggess ha equiy issuers are more likely o issue long erm deb. Furhermore, we see an increasing rend in oal accruals. The ime series average of TACC for ne repurchasers is no saisically significan, for ne issuers is 0.184, respecively, while he spread is (= ). From he las wo rows, we also see ime series averages of CACC han of NCACC for equiy repurchasers and dividend paying firms are no saisically significan, while for equiy issuers are 0.05 and 0.134, respecively. The spread of CACC and NCACC is (=-11.6) and (=-17.64), respecively. Overall, resuls from he las wo rows reveal ha firms wih high values of ne equiy financing are more likely o have high invesing capial accruals raher han working capial accruals. In Panel C of Table 1, we provide characerisics of porfolios formed on he magniude of ne deb financing aciviies. The ime series averages of DEBT for firms ha repay and issue deb are and -0.65, respecively, while he spread is 0.41 (=33.87). Noe also, 10

11 ha firms in he exreme deciles issue equiy, while in oher deciles are no engaged in equiy financing aciviies. This finding, suggess possible refinancing aciviy whereby he proceeds from equiy issuers could be used o repay deb. Resuls also reveal ha firms are more likely o engage in long erm han shor erm deb financing aciviies. The spread of SDEBT and LDEBT is 0.11 (=37.51) and 0.89 (=5.48), respecively. Furhermore, we see an increasing rend in oal accruals, working capial accruals and invesing capial accruals. The ime series averages of TACC for firms ha repay and issue deb are and 0.30, respecively, while he spread is (=-35.09). From he las wo rows, we also see similar in absolue value ime series averages of CACC han of NCACC for firms ha repay deb. On he oher hand, he ime series averages of CACC and NCACC for deb issuers are and 0.1, respecively. The spread of CACC and NCACC is (=-34.1) and (=--.75), respecively. Overall, resuls from he las wo rows reveal ha firms wih high values of ne deb financing are more likely o have high invesing capial accruals raher han working capial accruals. In Panel D of Table 1, we provide characerisics of porfolios formed on he magniude of ne shor erm deb financing aciviies. The ime series averages of LDEBT for firms ha repay and issue deb are and -0.1, respecively, while he spread is 0.7 (=4.1). Noe also, ha firms in exreme deciles are boh equiy and long erm deb issuers. This finding, sugges possible refinancing aciviy whereby he proceeds from equiy and long erm deb issuers could be used o repay shor erm deb. On he oher hand, firms in oher deciles are no engaged in equiy financing aciviies. Furhermore, we see an increasing rend in oal accruals, working capial accruals and invesing capial accruals. The ime series averages of TACC for firms ha repay and issue shor erm deb are and 0.16, respecively, while he spread is (=-3.96). From he las wo rows, we also see ha he ime series averages of CACC han of NCACC for shor erm deb issuers are and 0.086, respecively. On he oher hand, for firms ha repay shor erm deb he ime series average of CACC is -0.08, while of NCACC is no saisically significan. The spread of CACC and NCACC is (=-31.81) and (=-16.07), respecively. As such, resuls from he las wo rows ha working capial accruals are more likely o be associaed wih shor erm deb repaymen. Panel E of Table 1, repors characerisics of porfolios formed on he magniude of ne long erm deb financing aciviies. The ime series averages of SDEBT for firms ha repay and issue deb are and -0.38, respecively, while he spread is (=30.5). Noe also ha firms in exreme deciles are boh equiy issuers, a finding ha is indicaive of possible refinancing aciviy whereby he proceeds from equiy issues could be used o repay long erm deb. Similarly, we find shor erm deb repaymen for long erm deb issuers and shor erm deb issues for firms ha repay long erm deb. Furhermore, we see an increasing 11

12 rend in oal accruals, working capial accruals and invesing capial accruals. The ime series averages of TACC for firms ha repay and issue deb are and 0.64, respecively, while he spread is (=-7.9). From he las wo rows, we also see higher in absolue value ime series averages of NCACC han of CACC for firms ha boh issue and repay long erm deb. In paricular, he ime series averages of CACC and NCACC for firms ha repay long erm deb are and -0.04, respecively, while for firms ha issue long erm deb are and 0.1, respecively. The spread of CACC and NCACC is (=-.4) and (=-3.), respecively. Overall, resuls from he las wo rows reveal ha firms wih high values of ne long erm deb financing are more likely o have high invesing capial accruals raher han working capial accruals. Addiionally wih porfolio characerisics, cross secional correlaions beween exernal financing and accrual measures are compued each year. Mean correlaion over years are repored in Panel F of Table 1. From he firs row XFIN is highly correlaed wih boh EQUITY and DEBT. As such, boh EQUITY and DEBT represen significan sources of variaion in XFIN. Noe, ha XFIN and DEBT are more highly correlaed wih LDEBT han wih SDEBT, suggesing ha firms are more likely o be engaged in long erm deb han shor erm deb financing aciviies. Furhermore, he correlaion of XFIN, DEBT and LDEBT wih TACC is high and similar wih NCACC. On he oher hand, XFIN, DEBT and LDEBT are less correlaed wih CACC. Noe also, ha EQUITY is similarly correlaed wih TACC and NCACC and less wih CACC. This finding indicaes ha he close relaion of he exernal financing anomaly and he accrual anomaly may be aributable o invesing capial accruals. However, SDEBT is similarly correlaed wih TACC and CACC and less wih NCACC. 4. Sock Reurns Tess from Exernal Financing Porfolios. In his secion, we invesigae he performance of porfolios based on he magniude of exernal financing measures. As in he previous secion, we rank firms annually on each measure, allocae hem ino en equal-sized porfolios (deciles) based on hese ranks and hen compue heir fuure raw and size-adjused reurns. In Panel A of Table, we repor he ime series averages of raw reurns for each porfolio based on exernal financing measure, along wih heir associaed -saisics (in parenhesis). We also repor he ime series averages of reurns for hedge sraegies consising of a long (shor) posiion in he lowes (highes) decile. Saring wih XFIN, we see ha raw reurns for ne repurchasers and ne issuers are 0. (=5.36) and 0.09 (=1.978), respecively. A rading sraegy on XFIN generaes a raw reurn of abou 0.11 (=4.801). Turning o EQUITY, he raw reurn for equiy issuers and dividend paying firms is (=5.858), for equiy issuers is 0.1 (=.013), while for he hedge 1

13 sraegy is (=3.001). Firms ha repay deb have a raw reurn of abou 0.01 (=4.335), firms ha issue deb (=.501), while he hedge raw reurn for DEBT is 0.09 (=4.86), respecively. From a closer look o deb financing proxies, we see ha SDEBT and LDEBT hedge sraegies generae raw reurns of abou (=.81) and (=5.51), respecively. Thus, boh forms of ne deb financing aciviies are negaively relaed wih fuure sock reurns. However, he relaion is sronger for ne long erm financing aciviies. Panel B of Table presens ime series averages of size-adjused reurns for porfolios and hedge sraegies based on he magniude of exernal financing measures. From he firs column, we see ha he size-adjused reurn for ne repurchasers is (=4.314), for ne issuers is (=-3.339), while for he hedge sraegy on XFIN is (=4.8). Turning o EQUITY, we see ha size-adjused reurns for ne repurchasers and ne issuers are (=4.58) and (=-.054), respecively, while for he hedge sraegy is (=3.53). Furher, firms ha repay deb have a size-adjused reurn of abou 0.04 (=.98), firms ha issue deb (=-.685), while he hedge size-adjused reurn for DEBT is 0.08 (=4.61), respecively. Noe, ha Bradshaw e al. (006) repor similar size-adjused reurns for XFIN, EQUITY and DEBT. Turning o deb financing proxies, we see ha hedge sraegies on SDEBT and LDEBT generae reurns of abou (=3.016) and (=4.7), respecively. As such, he negaive relaion beween sock reurns and ne long erm financing aciviies is sronger han wih ne shor erm deb financing aciviies. Noe ha all sraegies are found profitable in he grea majoriy of years of our sample period. Resuls from our sock reurn ess in Panel A and B of Table indicae posiive raw and size-adjused reurns for hedge rading sraegies on exernal financing measures. However, as argued by Fama (1998) a problem in hese ess is ha all models of expeced reurns are incomplee descripions of he sysemaic paerns in average reurns during any sample period. As a resul, sock reurn ess are always conaminaed by a bad model problem. In order o check he robusness of resuls from our sock reurn ess, we apply he saisical arbirage es ha is designed by Hogan e al. (004) o hedge sraegies on all exernal financing measures. This es circumvens he bad model problem of sock reurn ess since i is no coningen upon a specific model for marke reurns. By definiion a rading sraegy ha consiues saisical arbirage opporuniies mus have a zero iniial cos (self financing), posiive expeced discouned profis, a probabiliy of a loss converging o zero and a ime-averaged variance converging o zero if he probabiliy of a loss does no become zero in finie ime. In economics erms, he las condiion associaed wih he imeaveraged variance implies ha a saisical arbirage opporuniy evenually produces riskless incremenal profi, wih an associaed Sharpe raio increasing monoonically hrough ime. 13

14 The zero iniial cos (self financing) condiion in hese ess is enforced by invesing (borrowing) rading profis (losses) generaed by each rading sraegy a he risk free rae. Specifically, ime series of annual hedge (raw) reurns RET ( i ) are firs generaed from accruals and value/growh sraegies. Then, he rading profis V ( i ) of each rading sraegy accumulae a he risk free rae r ( i ) o yield cumulaive rading profis (wih ( ) ( ) ( ) ( ) ( ) r i 1 i i i 1 V = ): 0 0 V = RET + e V (4) This cumulaive rading profi is hen discouned each period by i 1 Σ r n ( ) e = i o consruc discouned cumulaive rading profis v( i ) for each rading sraegy. Le ( ) ( ) vi = v i v i 1, denoe he incremens of he discouned cumulaive profis wih mean µ, growh rae of mean θ, sandard deviaion σ and growh rae of sandard deviaion λ. Assume also ha he incremens of he discouned cumulaive profis evolve according o he following sochasic process: v i θ λ = µ + σ i z (5) i where i=1,,..n, i i z are. i. d N( 0,1) i random variables wih 0 0 = vi z, ( ) v 0 and v0 are equal o zero. Under he above assumed sochasic process, he discouned cumulaive profis v are disribued as n n n θ λ v( n ) = vi ~ N µ i, σ i i= 1 i= 1 i= 1 (6) and have he following log likelihood funcion log L,,, v log i v i σ i n n λ θ ( µ σ θ λ ) = ( σ ) ( ) λ i µ (7) i= 1 i= 1 The parameers µ,θ,σ, λ can be esimaed hrough he maximum likelihood esimaion mehod and he associaed score equaions are provided in he appendix. Then, assuming ha θ = 0, one can conduc consrain mean ess of saisical arbirage. In paricular, under hese ess a rading sraegy generaes saisical arbirage wih1 α percen confidence if he following condiions are saisfied 11 : H1: µ > 0 and H: λ < 0. The firs hypohesis ess wheher he mean annual incremenal profi of a rading sraegy is posiive (second condiion for saisical arbirage) and he second, wheher is imeaveraged variance decreases over ime (fourh condiion of saisical arbirage). The wo parameers are esed individually wih he Bonferroni inequaliy accouning for he combined 11 See in he appendix he appropriae condiions for saisical arbirage under he unconsrain mean ess and in Hogan e al. (004) for furher deails on heir differences. 14

15 naure of he hypohesis es. Noe, ha sandards errors for he above parameers may be exraced from he Hessian marix o produce he required corresponding p-values. 1 From he repored resuls on Panel C of Table we see ha hedge sraegies on exernal financing and accrual measures consiue saisical arbirage opporuniies a he 1% level. Only, he sraegy on ne shor erm deb financing is found o survive he saisical arbirage es only a he 5% level. Noe, ha if one agrees ha he noion of saisical arbirage is incompaible wih marke equilibrium, and by inference, marke efficiency (Jarrow 1988, chaper 19), hen our evidence suppors exising behavioral explanaions o inerpre he role of accruals on he predicabiliy of sock reurns following exernal financing aciviies. 4.3 Sock Reurns Tess from Ineraced Porfolios on Ne Exernal Financing and Accrual Measures So far, he exernal financing anomaly has been examined independenly from he accrual anomaly. In his secion, we invesigae he role of accruals on he predicabiliy of sock reurns following exernal financing aciviies by considering conrol hedge and join hedge sraegies. 13 To implemen hese wo-dimensional sraegies, each year firms are sored independenly on exernal financing and accrual measures, and allocaed ino hree groupporfolios: he boom 0 percen (Porfolio 1), middle 60 percen (Porfolio ), and op 0 percen (Porfolio 3). As in he previous secions, porfolios are held for one year and hen rebalanced, while we require a leas a four-monh gap beween he porfolio formaion monh and he fiscal year end. We hen focus, on he resuled inersecions 14 from he above menioned sors. Under a conrol hedge sraegy we assess wheher he exernal financing effec survives, afer holding he accrual effec consan. Under a join hedge sraegy we assess wheher he combinaion of hese effecs, generaes an indicaor ha is significanly beer han eiher one effec separaely. Table 3 repors he size-adjused reurns for simple porfolios based on he magniude of exernal financing measures and heir inersecions wih porfolios based on he magniude of oal accruals, along wih heir associaed -saisics (in parenhesis). Noe ha he hedge size-adjused reurns for he uncondiional sraegies on XFIN, EQUITY and DEBT are 1 Noe ha hese hypoheses are he economic hypoheses for he presence of saisical arbirage. The saisical hypoheses under esing are µ = 0 and λ = 0 ha correspond o absence of arbirage. 13 Reinganum (1981), Jaffe e al. (1989), Greig (199), Hong e al. (000), Collins and Hribar (00) and Desai e al. (004) have used his approach o address relaed quesions. In supplemenary ess, we perform analysis by considering wo-dimensional sraegies from sequenial sors on exernal financing and accrual measures and find qualiaive similar resuls. 14 Using group analysis leads o lower sandard errors in -saisics for hedge reurns across wodimensional sraegies han decile analysis. This approach has been also used by oher sudies in he accouning and he finance lieraure. However, he resuls are qualiaively similar wih decile analysis. 15

16 0.09 (=5.564), (=.548) and 0.07 (=7.389), respecively. From a closer look o deb financing proxies, we see ha SDEBT and LDEBT hedge sraegies generae reurns of abou (=4.1) and (=5.17), respecively. From Panel A, we see ha he sraegy on XFIN is no profitable across firms wih low and high levels of TACC. Turning o Panel B, we see ha he sraegy on EQUITY generaes insignifican size-adjused reurns across firms wih low and medium levels of TACC. Resuls on Panel C, reveal ha he sraegy on DEBT is no profitable across firms wih low levels of TACC and earns negaive size-adjused reurns across firms wih high levels TACC. Furhermore, from Panels D and E we see ha size-adjused reurns o sraegy SDEBT are compleely subsumed by TACC, while he sraegy on LDEBT is no profitable across firms wih low and high levels of TACC. Noe also, ha he performance of hedge sraegies ha combine informaion on boh exernal financing measures and oal accruals is indisinguishable o ha of an uncondiional sraegy on oal accruals. As such, consisen wih prior evidence in he accouning lieraure, our findings from Table 3 indicae ha he anomalies on exernal financing aciviies are relaed wih he anomaly on oal accruals. Table 4 repors he size-adjused reurns for simple porfolios based on he magniude of exernal financing measures and heir inersecions wih porfolios based on he magniude of working capial accruals, along wih heir associaed -saisics (in parenhesis). From Panel A, we see ha he sraegy on XFIN is profitable across all firms regardless heir exposure o CACC. Similar evidence is found from Panels B, C and E for he sraegies on EQUITY, DEBT and LDEBT, respecively ( EQUITY on he middle CACC porfolio is he only excepion). However, resuls on Panel D show ha he sraegy on SDEBT is no profitable across firms wih medium and high levels of CACC. Furhermore, findings from all panels reveal ha he generaed size-adjused reurns from hedge sraegies ha combine informaion on boh exernal financing measures and working capial accruals are significanly higher han hose obained from each measure in isolaion. Overall, our findings from Table 4 sugges ha he anomalies on exernal financing aciviies are unrelaed wih he anomaly on working capial accruals, firs documened by Sloan (1996). However, our evidence on Panel D is suggesive of a significan role for working capial accruals on he predicabiliy of sock reurns following shor erm deb financing aciviies. Table 5 repors he size-adjused reurns for simple porfolios based on he magniude of exernal financing measures and heir inersecions wih porfolios based on he magniude of invesing capial accruals, along wih heir associaed -saisics (in parenhesis).from Panel A and C, we see ha he sraegies on XFIN and DEBT are no profitable across firms wih high levels of NCACC. Turning o Panel B, we see ha he sraegy on EQUITY generaes insignifican size-adjused reurns across firms wih low levels of NCACC. Resuls on Panel D and E, reveal ha he sraegies on SDEBT and LDEBT are no profitable 16

17 across firms wih low and high levels of NCACC. Furhermore, from all panels is found ha ha he generaed size-adjused from hedge sraegies ha combine informaion on boh exernal financing measures and invesing capial accruals are no significanly higher ha hose from an uncondiional sraegy on invesing capial accruals. These findings imply he anomalies on exernal financing aciviies are relaed wih he anomaly on invesing capial accruals, firs documened by Richardson e al. (005). Overall, our evidence from Table 5 indicaes ha he relaion of he anomalies on exernal financing aciviies and accruals is more likely o be driven from invesing capial accruals. 4.4 Regressions on Ne Exernal Financing and Accrual Measures. In his secion, we esimae Fama - MacBeh (1973) regressions of raw sock reurns on exernal financing measures and accrual measures 15, afer conrolling for size and book o marke raio, and repor he ime series averages of he resuling parameer coefficiens. The repored -saisics (in parenhesis) are based on he means and sandard deviaions of he parameer coefficiens obained in he annual cross secional regressions. To ensure ha resuls are no driven from exreme observaions we repea regressions for wo subsamples. To form hese subsamples, we firs divide he enire sample across he accrual dimension so ha one half conains predominanly low accrual firms and he oher predominanly high accrual firms. Then, we idenify issuers and repurchasers in each of hese groups. Based on hese pariions he firs subsample (overlap subsample) conains low accrual firms ha are also repurchasers (firms wih lower han mean accrual and exernal financing measures) and high accrual firms ha are also issuers (firms wih higher han mean accrual and exernal financing measures). The second subsample (nonoverlap subsample) conains low accrual firms ha are also issuers (firms wih lower han mean accrual measures and higher han mean exernal financing measures) and high accrual firms ha are also repurchasers (firms wih higher han mean accrual measures and lower han mean exernal financing measures). In his way, we invesigae he role of accruals on he predicabiliy of sock reurns following exernal financing aciviies. If accruals do no have an imporan role, hen he exernal financing effec should be srong for he full sample and for boh subsamples. Before discussing our resuls, noe ha from unrepored regressions we found negaive and saisically significan coefficiens on all exernal financing measures, uncondiional on 15 The regression approach imposes linear srucure on he relaion beween reurns and he variable under invesigaion, even hough he relaion may be non-linear. To conrol for poenial non-lineariies and ensure ha resuls are no driven from exreme observaions, we follow Desai e al. (004) and express variables as porfolio decile ranking. In paricular, each year we sor firms independenly ino nine deciles (0,9) based on exernal financing and accrual measures and divide he decile number by 9 so ha each firm-year observaion relaed o hese variables akes a value ranging beween 0 and 1. 17

18 accrual measures. 16 In Table 6, we repor resuls from regressions of raw sock reurns on each exernal financing measure afer conrolling for oal accruals. From Panel A, i is found ha when boh exernal financing measures and oal accruals are included in he regressions, he coefficiens on XFIN, EQUITY, DEBT, SDEBT and LDEBT are no saisically significan, while he coefficien on TACC is negaive and saisically significan. Similar resuls are repored in Panels B and C for he overlap and nonoverlap subsample, respecively. As such, consisen wih evidence in Cohen and Lys (006), hese findings indicae ha he exernal financing anomaly no longer persiss, once we conrol for oal accruals. Table 7, provides resuls from regressions of raw sock reurns on each exernal financing measure afer conrolling for working capial accruals. From Panel A we see ha when boh exernal financing measures and working capial accruals are included in he regressions, he coefficiens on all measures are negaive and saisically significan. Turning o Panels B and C, we see similar resuls for XFIN, EQUITY, DEBT, LDEBT and CACC across he overlap and nonoverlap subsample, respecively. These findings imply ha he predicive power of hose exernal financing measures for fuure reurns is unrelaed o ha of working capial accruals. However, for he overlap subsample he coefficien for SDEBT is no found saisically significan, while for he nonoverlap subsample is found negaive and saisically significan. This finding indicaes ha here is a relaion beween he anomalies on shor erm deb financing aciviies and working capial accruals. In Table 8, we repor resuls from regressions of raw sock reurns on each exernal financing measure afer conrolling for invesing capial accruals. Resuls from Panel A reveal ha when boh exernal financing measures and invesing capial accruals are included in he regressions, he coefficiens on XFIN, DEBT, SDEBT and NCACC are negaive and saisically significan, while on EQUITY and LDEBT are saisically insignifican. However, from Panel B ha provides resuls for he overlap subsample we see ha he predicive power of all exernal financing measures for fuure sock reurns is closely relaed and subsumed by ha of invesing capial accruals. Similar resuls are repored on Panel C for he nonoverlap subsample, excep XFIN. These findings sugges ha here is a srong relaion beween he anomalies on exernal financing aciviies and invesing capial accruals. Overall, he resuls from regression analysis confirm prior evidence from porfolio level analysis ha he srong relaion of he anomalies on exernal financing aciviies and accruals is more likely o be driven from invesing capial accruals. In oher words, invesing capial accruals is a key in undersanding he exernal financing anomaly. Working capial accruals seem o play an imporan role only on he predicabiliy of sock reurns, following shor erm deb financing aciviies. 16 Similar resuls are repored in Bradshaw e al. (006) and Cohen and Lys (006). 18

19 5. Conclusions An exensive body of empirical work in finance and accouning sudies documens a negaive relaion beween ne exernal financing aciviies and fuure sock reurns. However, Cohen and Lys (006) show ha ha he exernal financing anomaly no longer persiss, afer conrolling for oal accruals. Based on Richardson e al. (005) accruals can be divided o working capial accruals ha represen growh in ne working capial and invesing capial accruals ha represen growh in ne invesing capial. The naure of differen informaion in accrual componens for he predicabiliy of sock reurns following exernal financing aciviies has no been horoughly explored, however. This issue moivaes wha we do in his paper. Our findings can be summarized in wha follows. We show ha firms wih high values of ne exernal financing are more likely o have high invesing capial accruals raher han working capial accruals. We also show ha hedge sraegies on ne changes in enire exernal financing, equiy financing ransacions and deb financing ransacions generae posiive raw and size-adjused sock reurns. The sraegies are also found o consiue saisical arbirage opporuniies. Similar resuls are found when we disinguish beween ne shor erm and ne long erm deb financing aciviies. However, he effecs are found sronger for ne long erm deb financing aciviies. We also show ha he srong relaion of he anomalies on exernal financing aciviies and accruals is more likely o be driven from invesing capial accruals. Working capial accruals seem o have an imporan role only on he predicabiliy of sock reurns, following shor erm deb financing aciviies. Overall, our evidence is more likely o be consisen wih invesor s failure o recognise opporunisic earnings managemen and/or agency relaed overinvesmen associaed wih invesed capial. However, our findings do no necessarily rule ou a risk based explanaion. I is possible ha an omied risk facor relaed o earnings qualiy and/or capial invesmen may be he underlying culpri for he srong relaion beween anomalies on exernal financing aciviies and accruals. I would be more ineresing for fuure research o disenangle beween he impac of earnings managemen and invesmen on he exernal financing effec. 19

20 Appendix A. Parameers Esimaes for he Saisical Arbirages Tess The parameers µ,θ,σ, λ are esimaed from he following sysem of four equaions wih four unknowns: ( µ σ θ λ v) log L,,, µ ( v) : µ = n i= 1 n i= 1 v i i θ λ i ( θ λ ) n : σ n i 1 i λ = ( vi i ) log L µ, σ, θ, λ 1 1 σ = µ θ () ( µ σ θ λ v n n ) θ λ ( θ λ ) : vi log ( i) i = µ log ( i) i (3) log L,,, θ i= 1 i= 1 ( µ σ θ λ n n ) ( ) : σ log ( i) ( vi µ i θ = ) (4) log L,,, v log i λ i= 1 i= 1 i λ (1) Noe ha by assuming, θ = 0 and λ = 0 we ge he sandard MLE esimaors of he mean and he variance of he incremenal rading profis of each sraegy: 1 n vi n i= 1 µ = 1 n i and σ = ( v µ ) n i= 1 B. Unconsrain Mean Tes of Saisical Arbirage Under he unconsrain mean es, a rading sraegy generaes saisical arbirage wih1 α percen confidence if he following condiions are saisfied: H1: µ > 0 H: λ < 0 1 H3: θ > max λ, 1 wih he sum of p values for he individual ess forming an upper bound for he ype I error a. Noe ha by assuming θ = 0 he unconsrain mean es of saisical arbirage is reduced o a consrain mean es, while by assuming θ = 0 and λ = 0 -es. i is reduced o a single Finally, for he es of H o be well defined, we have o assume ha he parameer space for λ is he whole real live, alhough for v o have a well defined disribuion we need λ 0. 0

21 References Affleck-Graves, J., Miller, R The informaion conen of calls of deb: evidence from long-run sock reurns. Journal of Financial Research 6, Alford, W., Jones, J., Zmijewski, M Exensions and violaions of he sauory SEC Form 10-K filing requiremens. Journal of Accouning & Economics, 17, Anderson, C., Garcia-Feijoo, L Empirical evidence on capial invesmen, growh opions, and securiy reurns. Journal of Finance, 61, Ball, R., Kohari, S., Robin, A The effec of inernaional insiuional facors on properies of accouning earnings. Journal of Accouning and Economics, 9, Ball, R., Robin, A., Wu, J Accouning sandards, he insiuional environmen and issuer incenives: effec on imely loss recogniion on China. Asia Pacific of Journal of Accouning and Economics, 7, Ball, R., Robin, A., Wu, J Incenives versus sandards: properies of accouning income in four Eas Asian counries and implicaions for accepance of IAS. Journal of Accouning and Economics, 36, Basu, S The conservaism principle and asymmeric imelines of earnings. Journal of Accouning and Economics, 4, Bradshaw, M., Richardson, S., Sloan, R The relaion beween corporae financing aciviies, analyss forecass and sock reurns. Journal of Accouning and Economics, 4, Bille, M.T., Flannery, M.J., Garfinkel, J The long-run performance of firms following loan announcemens, Universiy of Iowa working paper. Cohen, D. A., and T. Z. Lys, 006. Weighing he evidence on he relaion beween exernal corporae financing aciviies, accruals and sock reurns. Journal of Accouning and Economics, 4, Collins, D., Hribar, P. 00. Earnings-based and accrual-based marke anomalies: one effec or wo? Journal of Accouning and Economics, 9, Daniel, K., Timan, S Marke reacions o angible and inangible informaion, Journal of Finance, 61, Dechow, P Accouning earnings and cash flows as measures of firm performance: he role of accouning accruals. Journal of Accouning and Economics, 18, 3-4. Dechow, P., Kohari, S., Was, R The relaion beween earnings and cash flows, Journal of Accouning and Economics, 5, Dechow, P., Richardson, S., Sloan, R The persisence and pricing of he cash componen of earnings. Journal of Accouning Research, 46, Desai, H., Rajgopal, S., Venkaachalam, M Value-glamour and accruals mispricing: one anomaly or ow? The Accouning Review, 79,