Market Timing and the Debt-Equity Choice.

Transcription

1 Make Timing and he Deb-Equiy Choice. William B. Ellio *, Johanna Koëe-Kan 2, and Richad S. Wa 3 College of Business Adminisaion, Oklahoma Sae Univesiy, Sillwae, OK Faculy of Economics and Business Adminisaion, Vije Univesiei, Amsedam, The ehelands 3 College of Managemen, oh Caolina Sae Univesiy, Raleigh, C 27695

2 Make Timing and he Deb-Equiy Choice. Absac We es he make iming heoy of capial sucue using an eanings-based valuaion model ha allows us o sepaae equiy mispicing fom gowh opions and ime-vaying advese selecion; hus avoiding he muliple inepeaions of book-o-make aio. We find ha equiy make mispicing plays a significan, if no dominan, ole in he secuiy choice decision. Ou esuls ae obus o he inclusion of poxies fo ime-vaying gowh opions and alenae mehods of measuing misvaluaion. JEL classificaion: G32 Keywods: Capial sucue; Make iming; Secuiy choice; 2

3 Inoducion The equiy make iming heoy of capial sucue poposes ha manages ae able o idenify imes when equiy issuance is less cosly compaed o ohe ypes of exenal financing due o he make s ovevaluaion of he fim s sock. Successful iming of he equiy make lowes he fim s cos of equiy and benefis cuen shaeholdes a he expense of new shaeholdes. If manages ae able o ime he equiy make, poxies fo misvaluaion should be coelaed wih he iming of he secuiy issuance decision. Some aemps have been made o documen a elaion beween poxies fo valuaion and equiy issuance. Loughan and Rie (995, fo example, use pos issuance euns as an indiec poxy fo valuaion and documen geae equiy issuances duing peiods of elaively high make values, whee make values ae assumed o be negaively coelaed wih ex-pos euns. Moe ecenly, Bake and Wugle (2002 use he empoal vaiaion in book-o-make aios as a poxy fo valuaion and find ha leveage is posiively elaed o pas book-o-make aios. Bake and Wugle ague ha he pas paens of book-o-make imbed evidence of pas make misvaluaion. A key dawback of using book-o-make o measue misvaluaion is he muliple inepeaions of he aio. Indeed, his issue is aised by Bake and Wugle (p. 27 who concede ha while hey believe hei esuls ae consisen wih equiy mispicing in he pesence of iaional invesos and/o manages, hee ae alenaive inepeaions. Thei findings ae also consisen wih an exension of Myes (984 pecking ode model, which includes ime-vaying advese selecion coss (Lucas and McDonald, 990. Unde such a se up, fims will ime equiy issuances elaive o advese selecion

4 coss. If book-o-make measues vaiaions in advese selecion, equiy issues ae moe likely when book-o-make (and advese selecion coss is (ae low. McConnell and Sevaes (995 and Sulz (990 use book-o-make o measue gowh opions and find evidence ha fims wih many gowh opions issue equiy o miigae he undeinvesmen poblem ha aises fom geae leveage. In boh of hese alenaive explanaions, low book-o-make is inepeed as a poxy fo gowh opions o advese selecion coss, and fims issue equiy when book-o-make is low. In his pape we aemp o sepaae make iming effecs (caused by iaional picing fom he effecs of gowh opions and advese selecion (due o asymmeic infomaion. To examine he elaive impoance of gowh opions and mispicing, we decompose book-o-make ino a gowh opion componen and a mispicing componen and hen diecly examine he ole ha each plays in he secuiy choice decision. Ou mehod is in he spii of Rhodes-Kopf, Robinson, and Viswanahan (2005 who, in a sudy of equiy mispicing and meges, decompose book-o-make ino wo componens; he aio of (ininsic value o make pice and he aio of book value o (ininsic value. Rhodes-Kopf, Robinson and Viswanahan inepe he fis componen (value-o-pice as a measue of mispicing and he second componen (book-o-value as a measue of gowh oppouniies. Ou fis se of ess uses a valuaion model ha elies on he assumpion of pefec foesigh by manages. The advanage of his model is ha i inceases he sample size and poxies fo he manage s infomaion se a he issuance decision ime. While he pefec foesigh model allows fo he sepaaion of gowh opions and mispicing, i Auoe and Kovacs (2005 find evidence of a elaion beween ime vaying advese selecion and equiy issuances. 2

5 does no allow us o deemine whehe he mispicing is caused by iaionaliy of he make o asymmeic infomaion beween he make and manages. To ackle his issue, we epea ou ess using a valuaion model in which all he inpus ae publicly available a he ime of he issuance (specifically, analys eanings foecass. This allows us o sepaae iaional mispicing fom ime vaying advese selecion coss. We conduc ou pefec foesigh ess on a sample of 3,78 public equiy and 5,39 non-conveible deb issues. Consisen wih he make iming heoy, we find ha fims wih ovevalued equiy ae moe likely o issue equiy, while hose ha ae faily valued o undevalued, ae moe likely o issue deb. Using he Rhodes-Kopf e al. (2005 mehod we ae able o measue he elaive impoance of gowh opions compaed o mispicing. We find ha mispicing has a significan influence on he secuiy choice decision (consisen wih fims following make iming saegies based on iaional invesos o manages, even afe conolling fo gowh opions. Pevious wok has been unable o measue he elaive impoance of gowh opions and mispicing because book-o-make has lagely seved as a single poxy fo boh. Ou findings esolve he dual inepeaion of he esuls of Bake and Wugle (2002. Subsiuing analys foecas eanings fo ex-pos eanings leaves he esuls qualiaively unchanged. We inepe his as fuhe evidence of make iming in geneal and in paicula suggesive of inveso iaionally. Finally we also find simila esuls when we use alenaive valuaion measues o esimae mispicing. The pape poceeds as follows; Secion 2 discusses pevious eseach and develops he hypohesis, Secion 3 pesens he daa and mehod, Secion 4 pesens he esuls and obusness checks and Secion 5 concludes. 3

6 2 Lieaue eview and hypoheses developmen Unil ecenly, he fim s capial sucue choice has been modeled by wo main heoies. The pecking ode heoy, in which he fim issues he cheapes secuiy fis, usually implies ha he fim pefes inenal o exenal financing (Myes, 984; Myes and Majluf, 984 and he saic ade off heoy in which he fim makes gadual adjusmens ove ime owad an opimal age capial sucue. Recen sudies sugges ha he iming of deb and equiy issues plays an impoan ole in copoae financial policy. A suvey of 392 CFOs by Gaham and Havey (200 finds only limied suppo fo he pecking ode and saic ade off heoies, bu eveals ha CFOs claim o acively engage in make iming pacices. Bake and Wugle (2002 ae he fis o specifically addess how make iming affecs he fim s capial sucue by uilizing book-o-make aios as a poxy fo fim valuaion. They show ha high leveage fims aised capial duing peiods when hei book-o-make values wee high and low leveage fims aised capial when hei book-o-make aios wee low. They inepe hese esuls as evidence ha capial sucue is he cumulaive oucome of pas aemps o ime he equiy make, and offe wo possible explanaions fo hei findings. Fis, if advese selecion coss vay, eihe acoss fims o acoss ime, and ae posiively elaed o book-o-make, fims ae moe likely o issue equiy when book-omake and advese selecion is low. 2 This is an applicaion of he pecking ode heoy o a muli-peiod seing in which advese selecion coss vay and fims do no eve o hei capial sucue because he coss of no doing so ae small. Thei second explanaion assumes ha makes ae inefficien, which allows manages o ime he 2 Lucas and McDonald (990 and Koajczyk, Lucas and McDonald (99 model he iming of equiy issues when advese selecion coss vay acoss fims. Choe, Masulis, and anda (993 examine ime vaying advese selecion coss. 4

7 equiy make and issue equiy when hey peceive ha invesos ovevalue he fim. A hid explanaion (no offeed by Bake and Wugle is povided by McConnell and Sevaes (995 and Sulz (990 who pesen evidence ha fims wih high levels of gowh opions will fund hese gowh opions hough equiy issuances ahe han deb in ode o avoid deb hold up poblems. Thei wok pesens he second of he wo faces of deb in which deb can educe fim value by leading o undeinvesmen. In a moe ecen pape, Dima and Thako (2007 develop a heoy whee manages and invesos shae he same infomaion, bu have diffeing pios. The degee of ageemen beween manages and invesos abou he fuue pospecs of a new invesmen deemines how he secuiy issuance choice will affec he fim s pos-invesmen sock pice. In his model, a high sock pice associaed wih equiy issuance is inepeed as evidence of high ageemen beween manages and invesos abou he fuue pospecs of a new invesmen. Impoan o ou sudy, he pedicion of he ime vaying advese selecion cos hypohesis, he gowh opions hypohesis, and he make ageemen hypohesis ae boadly he same as ha fo he make iming hypohesis, ha is; ha low book-o-make fims will favo equiy issuances. The use of book-o-make as a measue of mispicing pesumes ha he book value of equiy is a easonable measue of fundamenal value. Howeve, as discussed ealie, a poion of he diffeence beween make pice and book value may be diven by ohe facos, such as gowh oppouniies. Rhodes-Kopf, Robinson and Viswanahan (2005 ecognize his poblem and develop a simple decomposiion of he book-o-make 5

8 aio in which book-o-make is viewed as he poduc of book-o-value and value-omake, i.e.: 3 B M B V = ( V M whee B is he book value of equiy, M is he make value of equiy and V is he fundamenal value of equiy. In hei se up, value-o-make (V/M measues mispicing by he make, while book-o-value (B/V measues gowh oppouniies. Pevious auhos have used value-o-make measues o esimae mispicing of socks, bu such measues ae subjec o he ciicism ha hey may be coelaed wih book-o-make and may be capuing gowh opions ahe han mispicing. By adding he book-o-value measue o ou models, we ae able o sepaae ou he mispicing and gowh opion componens of make-o-book. Decomposiion of book-o-make only addesses one pa of he muliple inepeaions poblem, in ha we sepaae gowh opions fom mispicing. Howeve, i does no diecly ackle he issue of asymmeic infomaion. Unde he pecking ode model, asymmeic infomaion leads fims o avoid exenal equiy financing. Fims do no issue equiy because he make will aionally e-pice he sock immediaely afe he issue because of he poenial lemons poblem. Fama and Fench (2002 ague ha he fac ha many fims do issue equiy (and do so in lage amouns is inconsisen wih asymmeic infomaion being a significan issue. Theefoe, in ode fo asymmeic infomaion o play a ole in he secuiy issuance decision we mus assume ha he fim s manages believe ha hey have supeio infomaion abou he fuue pospecs of he fim and ha he make will unde eac when he fim acs on his infomaion 3 These auhos acually use make-o-book ahe han book-o-make, bu we have esaed hei decomposiion o be consisen wih ou se up. 6

9 asymmey. This vesion of he asymmeic infomaion soy elies on inveso iaionaliy afe he issuance, as he fim may be coecly piced based on he make s uncondiional infomaion se. Disinguishing beween pue iaional mispicing and mispicing due o diffeing infomaion ses depends upon he mehod of compuing V, he fundamenal value of he sock. We use he esidual income model as a measue of fundamenal value. We explicily incopoae he gowh oppouniies ino he measue (in he fom of fuue abnomal eanings, ime-vaying make inees aes, and avoid, o some degee he poenial ha accouning convenions may no accuaely eflec fundamenal value. 4 Lee, Myes and Swaminahan (999 find ha he esidual income model has significan pedicive abiliy (ove 20% of he vaiaion in he eal eun of he Dow 30 and heefoe pefoms well as a measue of valuaion. They also find ha book-o-make aios alone explain only abou one hid of one pecen of he vaiaion in he eal eun in he Dow 30 socks ove a one o eigheen monh ime hoizon, and conclude ha booko-make aios povide lile economic pedicabiliy of sock euns. This esul is boadly consisen wih Kohai and Shanken (997 who show ha while book-o-make aios have some pedicive powe ove euns duing he 926 o 99 ime peiod, hei pedicive abiliy is subsanially educed in he 946 o 99 sub-peiod. We compue a fim s fundamenal value wih wo diffeen inpus. The fis is a pefec foesigh model ha uses ex-pos ealizaions of fuue eanings and epesens ou poxy fo he manage s opinion of value a he ime of issuance. The second vesion of he model uses analys s eanings foecass and hus conains only infomaion 4 The esidual income model is compued as he book value of equiy plus he pesen value of fuue eanings in excess of he equied ae of eun (vey simila o Economic Value Added. The specific popeies of he model and is abiliy o capue gowh oppouniies ae discussed in secion

10 available o he make a he ime of issuance. Thee ae hee poenial explanaions fo he pefec foesigh model poducing a valuaion diffeen fom he cuen sock pice. Fis, we canno ule ou he possibiliy of a misspecified model, and ha he model meely poduces a noisy esimae of he make pice. If his wee he case we would expec no pedicable elaion beween he mispicing measue and fim behavio. Second, he model may capue pue misvaluaion, whee he make has iaionally failed o value he sock coecly. Finally, if he pefec foesigh model accuaely poxies fo he manage s infomaion se, and does yield a diffeen value han he sock pice, i is possible ha asymmeic infomaion beween he fim and invesos is diving he valuaion wedge and causing he mispicing. To ackle he hid possibiliy, we use he analys foecas model which is fee of asymmeic infomaion beween he fim and invesos because i uses only publicly available infomaion. Confimaion of he pefec foesigh esuls by he analys foecas model implies ha he secuiy issuance decision is no diven by asymmeic infomaion, insead, and pehaps moe oubling; he mispicing is diven by invesos iaionally misvaluing socks. Wih he wo measues of ininsic value, elaive o make pice, we examine he choice of secuiies in ligh of he make iming hypohesis, while conolling fo exan heoies of capial sucue. We seek o deemine wha, if any, mispicing of equiy exiss acoss diffeen ypes of secuiy issuances, a he ime of issue. We sudy deb, and public equiy issuances. 5 Consisen wih he make iming hypohesis we expec fims o issue equiy when equiy is mos ovevalued, and when equiy is undevalued we expec he fim o issue deb. We model he secuiy issuance decision as a funcion of 5 Due o a vey limied sample size of companies issuing pivae equiy, we omi hese fom ou analysis. 8

11 mispicing and poxies fo facos ha pevious eseaches have documened as impoan deeminans of leveage. 3 Daa and mehod 3. Sample selecion The iniial sample consiss of all non-financial, U.S. fims ha issued public seasoned equiy, and all non-conveible deb (boh public and pivae, duing he peiod We exclude financial companies due o he highly egulaed envionmen in which hey opeae. Filing and issue daes and issue specific vaiables ae obained fom Secuiies Daa Company s (SDC Global ew Issues daabase. We equie ha all fims have elevan daa available on CRSP and COMPUSTAT. In he case of muliple deb issues ha occu wihin 0 days, we combine he poceeds of he issues, use he fis filing dae, and ea hem as one offeing. The valuaion model equies one yea of accouning daa pio o, and a leas hee yeas of accouning daa afe he announcemen dae. The iniial sample conains 9,72 secuiy issuances which ae divided ino 3,78 public equiy issues and 5,39 deb issues. 6 Table pesens he numbe of issues fo each ype of secuiy fo each yea of he sample. Toal issues vay fom a high of 678 in 992 o a low of 277 in Vaiables o measue poenial make iming The key vaiable equied o es he make iming heoy is a measue of value elaive o pice a he ime of secuiy issuance. Ohe sudies ha have examined make iming have elied on pos issue euns, make aciviy, issuance euns o book-omake aios as poxies fo mispicing. Boadly following Rhodes-Kopf, Robinson, and 6 Ou iniial lising of issues fom SDC oals 40,307 of which 7,329 ae pivae deb, 4,237 ae public deb and 8,065 ae public equiy. The combinaion of finding daa in Compusa, and he esicions imposed by he valuaion model subsanially educe he sample o ha epoed. 9

12 Viswanahan (2005, we decompose book-o-make ino wo componens, book-o-value and value-o-make. We use he esidual income model o esimae a fim s ininsic value, allowing us o moe diecly measue mispicing. The esidual income model is also used in a simila conex by D Mello and Shoff (2000, Jinda (2000, and Dong, Hishleife, Richadson and Teoh (2002. Lee, Myes, and Swaminahan (999 show ha value-o-pice aios, whee value is based on a esidual income valuaion model, have geae pedicive powe in foecasing fuue euns han book-o-make aios, eanings-o-pice aios, and dividend yields. The ininsic value is hen scaled by make pice o yield a aio ha, in heoy, should equal uniy when he fim is faily valued and is geae (less han one if i is undevalued (ovevalued. We implemen he model assuming pefec foesigh. This assumpion allows us o esimae he model fo a lage sample, bu as discussed above, does no peclude an asymmeic infomaion based misvaluaion soy. In obusness ess, we use also analys eanings foecass, and obain simila esuls, consisen wih asymmeic infomaion no being a dive of ou main esuls. Use of he value-o-pice aio has seveal advanages. Fis, we do no ely on a make model and hus avoid he issues elaed o he use of daily sock euns in even sudies. 7 Second, he secuiy issuance lieaue has no only documened abnomal pefomance in he sho-un, bu also in he long-un. The esidual income model allows us o capue boh, sho-un and long-un effecs. Thid, by assuming pefec foesigh by manages (hough he use of ex-pos daa we ae able o include a poxy fo manage s pivae infomaion abou he fuue eanings of he company. 7 See Bown and Wane (980, 985, Coado and Zivney (992, and MacKinlay (997 fo a deailed discussion on he use of daily sock euns in even sudies. 0

13 In addiion o he asymmeic infomaion issue, we also acknowledge ha use of ex-pos daa does no exemp us fom he possibiliy ha managemen may manipulae accouning daa and heefoe bias he esuls. We amelioae his poblem by scaling he valuaion aio by he aveage valuaion aio fo he subjec fim fom he wo yeas pio o he issue. This appoach capues he empoal misvaluaion of he fim. Thee is also a possibiliy ha he esuls fom he ex-pos eanings-based esidual income model (RIM ae diven by he choice of secuiy and heefoe may be endogenous. The analys foecas eanings-based RIM model amelioaes his poblem. Ohlson (990, 995 demonsaes ha he esidual income model is heoeically idenical o he dividend discoun model in deemining he value of equiy. We povide a poof of his equivalence in Appendix A. Howeve, he use of accouning numbes in he esidual income model makes he model easie fo pacical implemenaion han he dividend discoun model. In he esidual income model, equiy value esimaes ae based on fuue abnomal eanings. The basic model deemines he ininsic value by adding o book value he discouned expeced eanings in excess of nomal eun on book value. T T i ( E V B E [ X i Bi ] ( 0 = 0 ( 0 * TV (2 i= whee TV is calculaed as; TV = E0 T [( X T * BT ( X T * B ]/ 2 (3 E(V 0 is he value of he fim s equiy a ime zeo, B 0 is he book value a ime zeo (Compusa iem 60, is he cos of equiy, and E 0 (X i ae he expeced fuue eanings fo yea i a ime zeo (Compusa iem 8. Time zeo is he ime a he end of he fiscal yea immediaely peceding he file dae, and T equals wo yeas. All vaiables ae scaled by shaes ousanding (Compusa iem 25.

14 The esidual income model, calculaed as he sum of book value, he pesen value of wo yeas of abnomal eanings, and a eminal value, capues he long un gowh oppouniies in a way ha may no be immediaely obvious. Fis, fuue conibuions o value ae expessed as abnomal eanings (simila o he concep of economic value added. Fo he make in aggegae he mean abnomal eanings should be aound zeo. Fims wih non-zeo abnomal eanings will see hei abnomal eanings eve o zeo ove ime, as compeiive pessues educe hei abiliy o ean abnomal pofis. 8 The eminal value eas he las yea s abnomal eanings as a level, nominal pepeuiy, which is equivalen o pojecing he absolue value of hese abnomal eanings declining in eal ems. Fims wih gowh oppouniies should have posiive abnomal eanings and he model will capue hese beyond he wo yeas of he explici eanings used in he model. I is conceivable ha he model may undevalue fims wih no gowh oppouniies in he sho un, bu subsanial gowh oppouniies in he long em, so in unepoed esuls we also esimae he valuaion model using fou and five yeas of daa following he issue dae and geneae qualiaively simila esuls. The cos of equiy,, is esimaed using a single faco model on a fim by fim basis. We also esimae he cos of equiy using he Fama-Fench hee faco model, and geneae simila esuls, bu find ha he esimaes using his appoach ae fa noisie han hose fom he single faco model. The equiy isk pemium is esimaed as he geomeic diffeence beween he eun on he S&P 500 and he isk fee ae fom 926 o 2004 and is 6%. Ou esuls ae subsanaively unchanged if we use a olling isk pemium based on he geomeic diffeence beween he S&P 500 euns and he isk 8 Convesely fims wih negaive abnomal eanings will, ove ime see hese eanings eve posiively as managemen (o new managemen akes seps o coec opeaional deficiencies. 2

15 fee ae ove he 60 monhs pio o he issuance. We use he sho-em T-Bill ae as a poxy fo he isk-fee ae of inees. Lee e. al (999 epo ha boh he sho-em T- Bill aes and he long-em Teasuy bond aes ae useful poxies fo he isk fee ae. Howeve, esimaes of V 0 based on he sho-em T-Bill ae oupefom hose using he long-em T-bond aes because hey have a lowe sandad deviaion and a fase ae of mean evesion. The esimaed ininsic value of he sock a he ime of issue announcemen is hen scaled by he make value of he sock. Mispicing is measued as he value o pice aio: E( V 0 VP 0= (4 P0 whee VP 0 epesens he esimaed absolue mispicing a ime zeo, P 0 epesens he make pice of he sock one monh pio o he secuiy issue, and V 0 epesens he ininsic value of he sock a ime zeo. 9 We also measue mispicing elaive o he pevious yeas as shown in Equaion 5, whee VP - and VP -2 ae he value o pice aios fo he wo yeas pio o he issue. VP 0 VPs = ( (5 VP VP 2 2 If no mispicing is pesen, VPs should equal. A VPs less han, indicaes ha he fim is ove-valued a ime =0 elaive o he pevious yeas. A VPs geae han one indicaes ha he fim is unde-valued. In addiion o he pefec foesigh valuaion model, we also compue he esidual income model using analys foecass of fuue eanings fom FisCall. We use he las aveage foecas made pio o he issue dae fo he fiscal yea-end immediaely 9 As a obusness check we use he pio yea end sock pice and geneae quaniaively simila esuls, albei a lile less significan. 3

16 following he secuiy issuance as well as he yea-end following his foecas peiod. Fuue book values ae esimaed using he accouning clean suplus elaion, which saes ha book value equals pio book value plus ne income less dividends (as in Lee Myes and Swaminahan, 999. All ohe elemens of he model ae implemened in he same way as befoe. We efe o he value-o-pice aio esimaed using analys foecass as FVP. The ime seies valuaion meic esimaed in his manne is called FVPs. 3.3 A compaison of valuaion models As discussed ealie, we ague ha he value-o-pice measue, compued using he esidual income model, is a moe accuae measue of mispicing compaed o he book-o-make aio. Howeve, hese wo vaiables ae poenially coelaed as hey shae he same denominao. In Table 2 we pesen pai-wise coelaions of he vaious mispicing measues and book-o-make. The coelaion beween VP and book-omake (BTM is a posiive, and significan, The coelaion beween VPs and book-o-make is lowe, bu significan a The coelaion beween he ime seies pefec foesigh model (VPs and he analys foecas model (FVPs is posiive, as we would expec, bu faily small in magniude, a Recall ha he only diffeence beween hese wo meics is he daa souce of he eanings. This coelaion implies ha assumpions abou he use of eanings can have a significan impac on he valuaion model. 4 Resuls 4. Univaiae evidence of mispicing. Table 3 pesens summay saisics fo he wo secuiy issue ypes. A lowe (highe VP (value-o-pice aio implies equiy ovevaluaion (undevaluaion. In heoy, one would expec he VP aio o have a mean of, howeve, his is no he case 4

17 fo he sample as a whole, and neihe is i necessaily he case fo he univese of socks. The mean of VP is dependen on he choice of he equiy isk pemium used in he cos of capial compuaion. We use a fixed isk pemium (he long-un diffeence beween he eun on he S&P 500 and T-Bills; howeve duing mos of he ime peiod unde examinaion, he isk pemium implied by pices was lowe, and declining. Thus in a bull make, i is likely ha he isk pemium is ovesaed. As we ae examining he changes in he VP aio, he level of he isk pemium is no paiculaly impoan, and by assuming a sable isk pemium, we ae educing he effec of make wide misvaluaion. Ohe eseaches have encouneed his issue, fo example, Lee, Myes and Swaminahan (999 find a simila esul in hei implemenaion of he esidual income model and noe ha hey could scale he isk pemium o a value ha would equae o a VP aio equal o a he sa of hei sample, bu his would have no effec on he oveall esuls. In Table 3, we obseve ha he mean and median VP aios fo he equiy issues (0.674 and 0.44, especively ae lowe han hose fo he deb issues (0.874 and 0.75, especively, consisen wih moe ovevalued fims pefeing o issue equiy ahe han deb. This finding holds fo he ime seies measue of elaive misvaluaion, VPs, alhough o a lesse degee. Oveall, we find ha deb issues ae on aveage lage han equiy issues, and have moe angible asses. We find simila, if no moe ponounced esuls fo he analys foecas measues, FVPs and FVP. Howeve, as he obsevaions fo hese meics ae concenaed in he lae pa of he sample due o daa availabiliy, hey ae coinciden wih he bull make of he 990s. In Table 4 we pesen ess of he pecenage of ovevalued and undevalued fims in each issue ype, as well as coss issue compaisons of mispicing. Panel A pesens 5

18 Wilcoxon sign ank ess of he popoion of fims wih VPs< (elaively ovevalued and VPs> (elaively undevalued. Fo boh issue ypes, hee ae a geae popoion of fims ovevalued elaive o hei pio yeas, alhough as we discussed ealie, his is mos likely an aifac of a declining expeced isk pemium. Thee ae, howeve, coss secional diffeences beween issue ypes. Fo example, 75.53% of equiy issues ae elaively ovevalued, wheeas only 56.4% of deb issues ae elaively ovevalued. A he boom of Panel A, hese non-paameic esuls ae confimed by esing he diffeence of he means and medians beween deb and equiy issues (saisically significan in boh cases. In Panel B of Table 4, we examine he diffeence beween VP fo all deb issues and VP fo equiy issues. We find ha he equiy issues ae significanly ovevalued, elaive o deb issues, and again his finding is confimed in he diffeence of he means es a he boom of he panel. This esul also holds when we examine he medians using a sign ank es. In addiion o he ess pesened in Table 4, we pesen mean VP valuaion aios gaphically in Figue fo he nine yeas aound he secuiy issuance. Fims issuing equiy ae clealy ovevalued duing he issue yea, and he ovevaluaion deceases duing he subsequen wo yeas. This esul is consisen wih empiical evidence ha socks undepefom following he issuance of new equiy. Thee is a less clea paen suounding he deb issues, and hee does no appea o be any sysemaic decease in VP associaed wih he deb issue a yea zeo. 4.2 Mulivaiae ess of he secuiy issuance decision In a mulivaiae seing we es how well he mispicing measue pedics he likelihood ha he fim will issue a paicula ype of secuiy. Ou geneal appoach is o 6

19 use a binay vaiable o code issue ype, and hen examine he impac of ou mispicing meic on he issuance decision, in a famewok ha conols fo ohe well documened deeminans of secuiy choice. Table 5 pesens he coefficien esimaes fom he following logi egession model: 0 IssueType j [ β β valuaion measue] γ [ ]] = Φ (6 [ 0 conols whee IssueType j, akes he value of fo deb and he a value of 0 fo equiy, and he valuaion measue is eihe he absolue measue (i.e. VP o FVP o he ime seies elaive measue (i.e. VPs o FVPs. The conol vaiables ae descibed in he following sub-secion Conol vaiables In his secion we discuss he vaiables ha we employ o conol fo alenaive inepeaions of ou misvaluaion measue, as well as fo known deeminans of capial sucue. We ely heavily on he wok of Flanney and Rangan (2006 who in un build upon he wok of Rajan and Zingales (995, Hovakimian (2003, Hovakimian, Ople, and Timan (200, and Fama and Fench (2002. These vaiables ae chosen o conol fo he vaious facos ha may influence he fim s choice of secuiy. Size Lage fims end o have moe deb, ae moe anspaen, have lowe asse volailiy and sell enough deb so ha he fixed cos of deb is low. Fo any given issue, lage fims will moe likely issue deb. We expec a posiive coefficien on he size vaiable. We use he naual log of sales (Compusa iem 2 as a poxy fo size. 0 Using a nomal cumulaive densiy funcion (i.e., pobi yields qualiaively simila esuls. We log all he VP measues because hey ae highly skewed. A Shapio-Fancia W es daa ejecs he null of nomaliy fo VP a he % level. 7

20 Gowh opions Ofen, book-o-make has been used as a poxy fo gowh opions. Howeve, book-o-make can have ohe inepeaions, such as misvaluaion. In an aemp o amelioae he poblem associaed wih book-o-make, we use seveal alenaive measues of gowh opions. The aio of capial expendiues scaled by oal asses is compued using Compusa iem 28 divided by iem 6. Reseach and Developmen expense is also scaled by oal asses and is compued as R&D expense (iem 46 divided by oal asses (iem 6. Highe levels of boh aios, pesumably lead o moe gowh opions. Fims wih moe gowh opions may aemp o poec his anicipaed gowh by limiing leveage, and heefoe R&D and CapEx ae expeced o have a negaive coefficien. We also measue gowh opions using he book-o-value aio (BV, whee value is compued using he esidual income model and book is he book value of he fim s equiy. This appoach follows he mehod se ou in Rhodes-Kopf e al. (2005. To capue he ime seies vaiaion in his aio we scale each yea s obsevaion by he aveage of he wo pevious yea s aios in a manne idenical o ha used o esimae VPs in Equaion 5. This vaiable is called BVs and is compued as: BV 0 BVs 0= ( (7 BV BV2 2 We expec o find a posiive coefficien on he gowh opion poxy. Pofiabiliy Highe pofis will incease eained eanings, which mechanically educes leveage, howeve, highe pofis inceases he fims deb capaciy and in un, may esul in highe leveage. Theefoe, in ou model, he coefficien on pofiabiliy is 8

21 indeeminae. We measue pofiabiliy as EBITDA (iem 3 divided by oal asses (iem 6. Tangibiliy Fims wih moe inangible asses will end o pefe equiy, as i is difficul o use such asses o collaealize deb. Asse angibiliy is measued as he aio of fixed asses (iem 8 o oal asses (iem 6. A posiive coefficien in ou model will be consisen wih ou expecaions. Depeciaion Fims wih moe depeciaion have less need fo deb ax shields and will be moe likely o issue equiy. We expec a negaive elaionship beween depeciaion and he pobabiliy he fim issues deb. Depeciaion is measued as depeciaion expense (Compusa iem 4 divided by Toal Asses. Tage leveage If fims move hei leveage o a long-un age, fims below he age ae moe likely o issue deb. We capue age leveage hough wo vaiables. Leveage is compued as; Leveage = (Toal Asses-Book Equiy/Toal Asses (8 We also include he indusy median leveage fo he fim s Fama Fench (997 indusy classificaion compued in he same manne as above. This vaiable is called IndLeveage. We expec he sign on leveage o be negaive and he sign on IndLeveage o be posiive if fims wih leveage below he indusy aveage issue deb o move owads an indusy age. Cedi qualiy Fims wih aed deb have access o geae quaniies of deb a lowe coss, and end o be moe leveed. Theefoe, we expec a posiive coefficien in ou model. We 9

22 include a vaiable Raed which akes he value if he fim has aed deb unde Compusa iem 280, and zeo ohewise. Tax benefis Fims in a highe copoae ax backe benefi moe fom he deducibiliy of inees paymens, and heefoe ae moe likely o pefe deb o equiy. We expec a posiive coefficien on ax ae in ou model. The maginal ax aes wee geneously povided by John Gaham (and used in Gaham (996 and ohe subsequen wok as poxies fo he fim s maginal ax ae. 2 Inees aes Fims may espond o changes in he level of inees aes in hei secuiy issuance decision. Fo example, a educion in he cos of deb may induce fims o issue moe deb elaive o equiy. We poxy fo he change in he level of inees aes by including a vaiable Rae Change which measues he change in he AAA Copoae deb ae ove he las yea. We obain his daa fom he FRED II daabase a he Fedeal Reseve Bank of S Louis. 3 In addiion o he conols discussed, we also include yea dummies (unepoed in ou egessions o conol fo ime facos such as ime vaying equiy isk pemium. 4.3 Regession Resuls Table 5 pesens logi models fo he choice beween deb and equiy. Model pesens he simples model, in which lnvps is used as he valuaion measue. In his model, he coefficien fo lnvps is posiive ( and significan a he one-pecen level, indicaing ha fims ha ae elaively moe ove-valued, ae moe likely o issue equiy. Logi coefficiens can be inepeed in wo ways. Fis, exponeniaing he 2 hp://faculy.fuqua.duke.edu/~jgaham/axfom.hml 3 hp://eseach.slouisfed.og/fed2/ 20

23 coefficien yields he odds aio; a measue of he change in he odds of a deb issuance fo a uni incease in he independen vaiable. Fequenly, in applicaions such as ous, eseaches will sandadize his aio o measue he change in odds fo a one sandad deviaion shif in he independen vaiable of inees. In ou case, he sandad deviaion of VPs is quie lage and may no pesen a ealisic shif in valuaion fo a paicula fim. Insead we examine he odds aio fo a 0% shif in he valuaion measue i.e. an incease fom a mean of o., epesening 0% undevaluaion. In his case he change in lnvpts is ln(.= The odds aio fo his magniude change is heefoe e * =.092. Theefoe a 0% decline in valuaion will incease he likelihood of a deb issuance by abou 9%. A simila magniude change will occu fo 0% ovevaluaion. 4 A second appoach fo inepeing he economic significance of logi coefficiens is he maginal change in he pobabiliy of a deb issuance. In his case, we ae measuing he change in pobabiliy fom he condiional pobabiliy (i.e., ha se a he mean values of all he independen vaiables. This can be compued as: 5 E[ y / x] e = x e βx βx e e βx βx β (9 whee y is he dependen vaiable, x is he independen vaiable and β is he coefficien. Fo Model, when using he mean of lnvps (which equaes o a mean VPs of , his equaion evaluaes o , implying ha a 0. incease in lnvps (abou a 0% change in VPs will esul in a 2.3% incease in he pobabiliy of a deb issuance. 4 A 20% and 40% incease in VPs anslaes o a 7% and 3% incease especively in he likelihood of deb issuance. 5 See Geene (2002 fo a complee discussion. 2

24 Model includes wo gowh opions poxies, CapEx and R&D. The coefficien on CapEx is insignifican while R&D is significanly negaive; suggesing ha fims ha have moe gowh opions will favo equiy ove deb. Seveal of he conol vaiables ae also significan. Fo example, he posiive coefficien on ln(sales indicaes ha lage fims ae moe likely o issue deb, as ae fims wih moe angible asses, geae pofiabiliy, aed deb and highe maginal ax ae. The signs on he conol vaiables ae lagely as expeced and have been well documened by pevious auhos such as Hovokamian e al (200, Fama and Fench (2002, Koajczk and Levy (2003, Kayhan and Timan (2004 and Flanney and Rangan (2006. The significance of lnvps, even afe inclusion of he conol vaiables, indicaes he impoance of make valuaion as a faco in he deb-equiy decision. In Model 2 we include lnbvs, which measues changes in gowh oppouniies. Wih he addiion of lnbvs, he coefficien on he mispicing measue, lnvps, is a highly significan In his case a 0% incease in VPs anslaes ino a 5% incease in he likelihood of a deb issuance (i.e. e.5022*0.0953, o a 3.682% incease in he pobabiliy of a deb issuance. Again, a oughly symmeic paen exiss fo a decease in VPs. The gowh opions poxy, lnbvs, has a posiive and significan coefficien of.0305, which implies ha a 0% incease in his vaiable will anslae ino appoximaely a 0% incease in he likelihood ha he fim will issue deb. We do no diecly include book-o-make in he egession wih VPs o VP fo he simple eason ha, by consucion, BMs and BM ae highly coelaed wih VPs and VP, especively. Inclusion of highly coelaed covaiaes in a egession esuls in he usual sympoms of mulicollineaiy, mos noably lage sandad eos fo he 22

25 coelaed vaiables. 6 Since VP is a componen of book-o-make and he collinea componen is excluded fom he egession, he inepeaion of VP is aleed. This ges a he basic moivaion behind he Rhodes-Kopf e al. mehod namely o sepaae ou valuaion effecs and gowh opions Absolue Misvaluaion In Models 3 and 4 we epea he analysis of Models and 2, bu insead of measuing elaive valuaion using VPs, we measue misvaluaion wih he absolue valuaion meic, VP. The absolue valuaion meic compaes he ininsic value V wih he sock pice P, bu does no scale he measue by he pevious wo-yea s V/P aios. The absolue measue is a moe diec measue of misvaluaion in ha i compaes pice o ininsic value, bu i assumes ha in equilibium V should equal P. If hee ae biases in he model (pehaps in he isk pemium used, hen i could be he case ha fo many fims, V neve acually equals P, bu hese fims ae no necessaily mispiced. The esuls fo he absolue measue ae qualiaively simila o hose of Models and 2, alhough he significance of he coefficiens is a lile lowe. Using Equaion 9 o evaluae he economic significance of V/P in Model 3, we find ha fo a mean V/P of , a 0% change in V/P esuls in a oughly.% incease in he pobabiliy of a deb issuance. Fo Model 4, which incopoaes he lnb/v gowh opion vaiable, a 0% change in V/P esuls in a.8% incease in he pobabiliy of a deb issuance Residual income model using analys s eanings foecass. The esidual income model ha we have implemened so fa assumes pefec foesigh, in which fuue ealized eanings ae unbiased cuen peiod foecass of fuue expeced eanings. We jusify his appoach (which has been used by ohe eseaches 6 Indeed, inclusion of book-o-make in ou egessions gealy educes he significance of VPs o VP. 23

26 on he gounds ha we ae aemping o measue ininsic value using he infomaion se available o he manages of he fim. 7 Howeve, his appoach is subjec o wo ciicisms. Fis, o he exen ha fuue eanings may depend on he capial sucue decisions made oday, we may be inducing an elemen of endogeneiy ino ou eseach mehod. Specifically, fims ha issue equiy ahe han deb will pos highe fuue ne income, due o lowe inees coss. Howeve, his endogeneiy will bias agains finding ha equiy issues ae ovevalued as highe fuue eanings implies a highe V. The second ciicism is ha if manages do indeed have pivae infomaion, he valuaion wedge pesen in he pefec foesigh model may be diven by asymmeic infomaion and no iaional mispicing. One way o es he obusness of ou esuls o hese ciicisms is o employ analys foecass of eanings insead of acual ealized eanings. To his end, we eesimae he esidual income model using eanings foecass obained fom FisCall (ou appoach follows ha of Lee, Myes and Swaminahan, 999. We use he las consensus foecas made befoe he secuiy issuance fo each of he, 2 and 3 fuue yea-ends. Whee he yea 3 yea-end foecas is unavailable we esimae he gowh ae beween he and 2 yea foecas and hen we use his gowh ae o esimae he 3 yea foecas. FisCall daa is available beginning in 990, so ou ess in his secion cove a subse of he full sample. Fuhemoe, we only compue ou foecas esidual income model fo 7 I s possible ha manage s esimaes ae biased. The impac on he esuls would depend on whehe manages ae opimisic o pessimisic abou fuue eanings ealizaions. Assuming hey ae oveopimisic (as his seems mos likely, hen hei expecaion of V, call i V m will be highe han he V ha we esimae based on ealized EPS. Theefoe V m /P > V/P, so manages will hink ha hei sock is less ovevalued han ou model esimaes. In his case, make iming manages would be less likely o issue equiy. This bias would hen go agains he pedicions of ou model and seve o weaken ou esuls. Fuhe, when using he VPs measue, any manageial bias (so long as he manage s bias emains saic ove ime will be negaed, as he measue is scaled by pio yea s V/P aios. 24

27 he fim yeas in which we wee able o compue he pefec foesigh model hus avoiding any bias due o selecive coveage of he analys foecas daa. Table 6 pesens he esuls fom e-unning he models specified in Table 5, bu using analys foecass eanings o compue he valuaion model. In Models and 2 we measue misvaluaion in elaive ems using FVPs, while in Models 3 and 4 we measue misvaluaion as FVP. Again as wih he esuls fom Table 5 we find he mispicing meic is posiive and significan in all he models. I could be agued ha in an efficien make, one would no find any mispicing when analys foecass ae used in lieu of he ex-pos eanings. Howeve, ohes (Lee, Myes and Swaminahan, 999 have found simila esuls. We inepe hese findings as evidence ha he mispicing in he lage sample is no due o infomaion asymmey beween he fim s manages and invesos, bu due o make iaionaliy Alenaive valuaion measues If manages ae esponding o make valuaion of hei fim s equiy when hey make he decision o sell secuiies, we should obseve ha misvaluaion capued by ohe valuaion meics is also coelaed wih he secuiy issuance decision. 8 In Table 7 we exploe his issue fuhe by using pice-o-eanings aios, pice-o-sales aios and he sock pice un up as poxies fo ovevaluaion signals ha migh be eceived by manages. We expess boh he pice-o-eanings and pice-o-sales aios in absolue ems and in elaive ems in he same manne as employed fo VP. The pice-oeanings aio is he pice in he monh befoe he issue divided by he mos ecen yea end eanings pe shae befoe exaodinay iems (Compusa iem d8. 8 We would like o hank he efeee fo suggesing his exension. 25

28 The fis model in Table 7 uses he ime seies P/E aio, PEs which is compued as = PE 0 PEs 0 ( (0 PE PE2 2 whee PE 0 is he pice o eanings aio in he issue yea. As wih he ohe misvaluaion measues, we log all he valuaion aios, as hey ae heavily skewed. Logging he PE aio does esul in he loss of some obsevaions fo which eanings ae negaive, howeve he inepeaion of negaive pice-o-eanings aio is no clea. In Model he coefficien on PEs is significan and negaive implying ha an incease in he fim s PE aio elaive o pevious yeas means ha he fim is moe likely o issue equiy. A simila esul is found when we examine Model 2 which uses he ime seies pice o sales aio PSs. In Models 3 & 4 we eplace he ime seies measues of elaive misvaluaion wih he conempoaneous pice-o-eanings and pice-o-sales aios and again find ha highe absolue measues of hese aios make an equiy issuance moe likely. Finally in Model 5 we examine a un up in he sock pice (we acually use make value o conol fo sock splis and changes in shaes ousanding. The pice un up is compued as: = MV 0 MVs 0 ( ( MV MV2 2 whee MV 0 is he make value of he fim in yea. A un up in he value of he fim s equiy ove he pevious wo yeas is again moe likely o lead he fim o issue equiy, 26

29 consisen wih he manages inepeing his un up as a signal ha he fim s equiy is now ovevalued. 9 While he esuls of ou ess using alenaive valuaion aios confim he esuls using he esidual income model valuaion aios, hee ae ohe easons ha aionally explain an incease in all of hese alenaive measues. Fo example, an incease in PE aios could be due o a educion in he fim s sysemaic isk, an incease in expeced gowh o an incease in he opeaing efficiency of he fim. Pice o sales aios, while popula among equiy analyss, suffe he significan daw back ha hey can vay wih leveage. The advanage of he esidual income model appoach is ha i povides an explici poin esimae of he fim s value, ahe han a compaison wih jus a single dimension of he fim (such as eanings. 5 Conclusion We es he make iming heoy of capial sucue in a famewok ha avoids he dual inepeaion poblem of book-o-make, in which book-o-make measues gowh opions, asymmeic infomaion o iaional equiy mispicing. We employ an eanings based valuaion model (he esidual income model o diecly measue he fim s ininsic value. Using a sample of 9,72 secuiy issuances we find ha fims whose equiy is ovevalued, (i.e. he make value exceeds he ininsic value geneaed by he esidual income model ae significanly moe likely o issue equiy. In ode o diffeeniae beween he impac of mispicing and gowh opions in he secuiy issuance decision, we decompose book-o-make ino wo componens; 9 We do no include lnmvs in ou Table 5 egessions as a conol vaiable because lnmvs is highly coelaed wih lnvps (ho= Inclusion of lnmvs leads o mulicollineaiy and inceases he sandad eo of lnvps. In unepoed egessions, when his vaiable is added o model 2 of Table 5, he coefficien on lnvps falls o , bu emains saisically significan a he % level. 27

30 mispicing (value-o-pice and gowh opions (book-o-value. Afe conolling fo gowh opions, we find ha mispicing has significan incemenal explanaoy powe. To conol fo asymmeic infomaion ha may be pesen in he pefec foesigh model, we use a valuaion model based on analys foecass and find ou esuls ae emain obus o his alenaive specificaion. Finally, ou esuls ae obus o alenaive valuaion measues such as pice-oeanings aios, pice-o-sales aios and a measue of he sock pice un-up. We conclude ha mispicing of equiy plays a majo, if no dominan ole, in he secuiy choice decision. 28

31 29 Appendix A: Poof of he equivalence of he dividend discoun model and he esidual income model The dividend discoun model R D V ( ( 0 = = (A wheev 0 is he value of equiy a =0, D is he expeced dividend a, is he discoun ae, and R is he eminal value a =. The clean suplus elaion saes ha = B B X D (A2 whee X is he expeced eanings fo he peiod ending a, and B is he book value of equiy a. Subsiuing (A2 ino (A and doing some algeba R B B B X R B B B B X R B B B B X R B B X R B B X V ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( = = = = = = = = = = = = = = = since: ( ( ( = ( ( R B B B X V = = ( ( 0 0 as, assuming eanings ae gowing ino pepeuiy a a ae less han he cos of equiy, implies ha he eminal value and book value a = will be zeo. 0 0 ( ( B B X V = = This is he esidual income model as commonly implemened.

32 Refeences Auoe, D. and T. Kovacs, 2005, The pecking ode heoy and ime vaying advese selecion coss, woking pape: Viginia Tech. Bake, M., and J. Wugle, 2002, Make iming and capial sucue, Jounal of Finance, 57, -32. Bown, S., and J. Wane, 980, Measuing secuiy pice pefomance, Jounal of Financial Economics, 8, Bown, S., and J. Wane, 985, Using daily sock euns: The case of even sudies, Jounal of Financial Economics, 4, 3-3. Choe, H., R. Masulis, and V. anda, 993, Common sock offeings acoss he business cycle: Theoy and evidence, Jounal of Empiical Finance,, -3. Coado, C. J., and T. L. Zivney, 992, The specificaion and powe of he sign es in even sudy hypohesis es using daily sock euns, Jounal of Financial and Quaniaive Analysis, 27, Dima, A., and A. Thako, 2007, Why do fims issue equiy, Jounal of Finance, 62, D Mello, R., and P. K. Shoff, 2000, Equiy undevaluaion and decisions elaed o epuchase ende offes: An empiical invesigaion, Jounal of Finance, 55, Dong, M., D. Hishleife, S. Richadson, and S. Teoh, 2002, Does inveso misvaluaion dive he akeove make? woking pape: Ohio Sae Univesiy. Fama, E., and K. Fench, 997, Indusy coss of equiy, Jounal of Financial Economics, 43, Fama, E., and K. Fench, 2002, Tesing ade-off and pecking ode pedicions abou dividends and deb, Review of Financial Sudies, 5, -34. Flanney, M., and K. Rangan, 2006, Paial adjusmen owad age capial sucues, Jounal of Financial Economics, 79, Gaham, J. R., 996, Poxies fo he copoae maginal ax ae, Jounal of Financial Economics 42, Gaham, J. R., and C. R. Havey, 200 The heoy and pacice of copoae finance: Evidence fom he field, Jounal of Financial Economics, 60,

33 Geene, W. H., 2002, Economeic Analysis, 5 h Ediion, MacMillan, ew Yok. Hovakimian, A., T. Ople, and S. Timan, 200, The deb-equiy choice, Jounal of Financial and Quaniaive Analysis, 36, -24. Hovakimian, A., 2003, Ae obseved capial sucues deemined by equiy make iming?, woking pape: Bauch College. Jinda, J., 2000, Seasoned equiy offeings, ovevaluaion, and iming, woking pape. Kayhan, A., and S. Timan, 2004, Fims' Hisoies and Thei Capial Sucues, woking pape: Univesiy of Texas a Ausin. Kohai, S. P. and J. Shanken, 997, Book-o-make, dividend yield and expeced make euns: A ime seies analysis, Jounal of Financial Economics, 44, 2, Koajczyk, R. A., and A. Levy, 2003, Capial sucue choice: Macoeconomic condiions and financial consains, Jounal of Financial Economics, 68, Koajczyk, R. A., D. J. Lucas, and R. L. McDonald, 99, The effec of infomaion eleases on he picing and iming of equiy issues, Review of Financial Sudies, 4, Lee, C., J. Myes, and B. Swaminahan, 999, Wha is he ininsic value of he Dow? Jounal of Finance, 54, Loughan, T., and J. Rie, 995, The new issues puzzle, Jounal of Finance, 50,, Lucas, D., and R. McDonald, 990, Equiy Issues and Sock Pice Dynamics, Jounal of Finance, 45, MacKinlay, A. C., 997, Even sudies in economics and finance, Jounal of Economic Lieaue, 35, McConnell, J and H. Sevaes, 995, Equiy owneship and he wo faces of deb, Jounal of Financial Economics, 39, Myes, S., 984, The capial sucue puzzle, Jounal of Finance, 39, Myes, S, and Majluf,., 984, Copoae financing and invesmen decisions when fims have infomaion ha invesos do no have, Jounal of Financial Economics, 3, Ohlson, J., 990, A synhesis of secuiy valuaion heoy and he ole of dividends, cash flows, and eanings, Conempoay Accouning eseach, 6,

34 Ohlson, J., 995, Eanings, book values, and dividends in equiy valuaion, Conempoay Accouning eseach,, Rajan, R. G., and L. Zingales, 995, Wha do we know abou capial sucue: Some evidence fom inenaional daa, Jounal of Finance, 50, Rhodes-Kopf, M., D. T. Robinson and S. Viswanahan, 2005, Valuaion waves and mege aciviy: The empiical evidence, Jounal of Financial Economics, 77, Sulz, R., 990, Manageial disceion and opimal financing policies, Jounal of Financial Economics, 26,