A Valuation-Based Test Of. Equity Market Timing

Transcription

1 Deparmen of Finance Universiy of Melbourne Finance Research Essay A Valuaion-Based Tes Of Equiy Marke Timing Qian Zhang November 2007 i

2 Absrac Using a large sample of US firms beween 1976 and 2005, his paper examines he implicaions of equiy marke iming on capial srucure. Insead of using radiional marke-o-book raio which is facing muliple inerpreaion problems, his paper proxies equiy misvaluaion using price-value raio calculaed from he residual income model. Wih price-value raio, I consruc he equiy marke iming measure and find ha firms whose equiy is overvalued are more likely o reduce heir leverage hrough equiy issuance. In addiion, his impac of marke iming on capial srucure is found o be highly persisen. Overall, he evidence is consisen wih he marke iming heory, raher han he dynamic rade-off heory wih adjusmen coss. ii

3 Declaraion This essay is he sole work of he auhor whose name appears on he ile page. I conains no maerial which he auhor has previously submied for assessmen a he Universiy of Melbourne or elsewhere. To he bes of he auhor s knowledge, he essay conains no maerial previously wrien or published by anoher person excep where reference is made in he ex of he essay. Signaure of Suden iii

4 Acknowledgemens Many hanks are exended o all he academic saff and my fellow Honours sudens from he Deparmen of Finance a he Universiy of Melbourne. Thanks also o several ohers who provided valuable suppor hroughou he year. Special hanks o Dr. Xin Chang for his guidance, suggesions and infecious work ehic as well as some much needed words of encouragemen. iv

5 Table of Conens Main Body 2. Exising Capial Srucure Research Daa and Marke Timing Measure Daa and Conrol Variables Residual Income Model using Analyss Forecass Exernal Finance Weighed Average Price-Value Raio Summary Saisics Empirical Mehodology Deerminans of Annual Changes in Leverage Deerminans of Leverage Persisence Fama-MacBeh Regression Mehod Empirical Resuls Deerminans of Annual Changes in Leverage Deerminans of Leverage Persisence Robusness Check Conclusion Reference Table I: Summary Saisics of Firm Characerisics and Financing Decisions Table II: Deerminans of Annual Changes in Leverage and Componens Table III: Deerminans of Leverage Table IV: Deerminans of Cumulaive Changes in Leverage from he IPO Value Table V: Persisence Tes (Using 5-year-lagged marke iming measure) Table VI: Persisence Tes (Using 10-year-lagged marke iming measure) Table VII: Robusness Tes of Deerminans of Leverage Table VIII: Robusness Tes of Persisence effec from Marke Timing v

6 1. Inroducion Firms may fund heir invesmen projecs hrough inernally generaed funds, deb or equiy issuance. Wha deermines firms financing decisions is a very imporan area of sudy in he corporae finance lieraure. Tradiional heories of capial srucure ry o address his quesion using eiher rade-off heory or pecking order heory. In rade-off heory, firms have a arge capial srucure which is deermined by he marginal cos and benefi of deb. I predics ha firms adjus heir capial srucures around his arge leverage hrough ime in response o he emporary shocks. Under he pecking order heory, firms follow a financing hierarchy due o informaion asymmery beween managers and invesors inernally generaed funds, hen deb, and only as a las resor, exernal equiy financing. Recenly, anoher srand of lieraure has increasingly challenged boh radiional heories. The marke iming heory claims ha managers aemp o ime equiy marke by issuing shares a high prices and repurchasing shares a low prices. The exising empirical evidence on marke iming heory is sill inconclusive. The debae cenre on he proxy used o capure he exen of marke iming. Baker and Wurgler (2002) capure marke iming using he weighed-average of pas marke-o-book raios. In heir weighing scheme, he iming measure akes high values for firms ha raised exernal finance when he marke-o-book raio was high. Using high marke-o-book raio as a proxy for overvaluaion, hey argues ha firms ha raise more exernal finance when marke-o-book raio is high are more likely o be he firms ha ime he marke by issuing equiy. These firms should have a larger proporion of equiy in heir oal exernal finance. Furher, if firms have no arge 1

7 leverage raio, hey will allow he effecs of marke iming o accumulae and finally drive he firms capial srucure. Baker and Wurgler find ha, conrolling for firms invesmen opporuniies using curren marke-o-book, he curren leverage is negaively relaed o heir iming measure. They also find ha he resuling effecs on capial srucure are persisen and conclude he capial srucure is he cumulaive oucome of pas aemps o ime he equiy marke. However, Hovakimian (2006) has differen inerpreaions of he similar resuls. He conends ha Baker and Wurgler s iming measure conains informaion abou he firm s growh opporuniies no capured by he curren marke-o-book raio. Thus Baker and Wurgler s finding is consisen wih he fac ha firms end o use more equiy financing when heir pas growh opporuniies are high. Ali (2006) invesigaes he effec of marke iming in he conex of a major financing even, he iniial public offering. His marke iming measure is a ho marke dummy defined using he monhly IPO volume. He finds ha ho-marke IPO firms issue subsanially more equiy, and lower heir leverage by more, han cold-marke firms do. However, immediaely afer going public, ho-marke firms increase heir leverage by issuing more deb and less equiy relaive o cold-marke firms, suggesing ha he impac of marke iming is shor-lived. Liu (2007) uses insider rading aciviy as a measure of misvaluaion, and he evidence is more consisen wih a model of dynamic arge wih parial adjusmen. Ellio, Koëer-Kan and Warr (2007) employ an earnings-based fundamenal valuaion model o es he impac of marke iming on he firm s mehod of funding he financing defici. They find ha overvalued firms are more likely o issue equiy o fund heir 2

8 defici han undervalued firms. The effec of marke iming due o equiy misvaluaion is sill inconclusive. The objecive of his paper is wo-fold. Firs, i aims o consruc a more appropriae measure ha capures he essence of marke iming. In addiion, i examines he impac of equiy marke iming on firms capial srucure using he newly defined iming proxy. The iming measure is developed from he essence of marke iming heory, which saes ha managers ry o issue equiy when sock is overpriced and buy back shares when hey are underpriced. Then any es of he marke iming represens a join es of he heory iself and he mehod used o measure misvaluaion. The use of marke-o-book raio o es marke iming is fraugh wih difficulies. There are muliple inerpreaions of wha he marke-o-book raio capures: invesmen opporuniies, firm performance and misvaluaion. Despie hese muliple inerpreaions, he performance of marke-o-book raio as a measure of misvaluaion is relaively weak. 1 Unlike Baker and Wurgler (2002) who use marke-o-book raio o measure misvaluaion, I measure he misvaluaion explicily by esimaing he value of he firm s sock using he residual income model. The residual income model has been used exensively in he accouning lieraure. There is subsanial empirical and 1 La Pora (1996), Frankel and Lee (1998) find ha he premise ha high marke-o-book firms underperform low marke-o-book firms appears o be ime dependen. Kohari and Shanken (1997) find ha marke-o-book raios have some predicive power over he period, bu ha power is subsanially reduced during he sub-period. Lee, Myers and Swaminahan repor he evidence ha marke-o-book raios predic only abou 0.33% of he variaion in real sock reurns, and conclude ha marke-o-book is a weak measure of mispricing. 3

9 heoreical suppor exiss for is performance as a valuaion meric. 2 The inrinsic value is calculaed as he repored book value, plus an infinie sum of discouned fuure abnormal earnings under he residual income model. I scale his inrinsic value esimaes by he marke price o generae a clean, and easily inerpreed measure of mispricing. I hen consruc he iming measure as he exernal finance weighed average price-value raio (EFWAPV). By aking weighed average using equiy issuance as weighs, i akes high values for firms ha raised exernal finance when he price-value raio was high. Unlike Ellio, Koëer-Kan and Warr who use he lagged price-value raio, he iming measure in his paper picks ou he accumulaed hisorical marke iming effec on capial srucure. I is also specific o each individual firm s misvaluaion, raher han he aggregae marke measure Ali uses in his sudies. Using his more appropriaely defined marke iming measure, he es in his paper should beer capure he effec of equiy marke iming on capial srucure. I examine he marke iming effec using 4007 US firms from COMPUSTAT over he period 1976 o The inrinsic value is compued using analys forecass obained from I/B/E/S. All ess in his paper are Fama-MacBeh (1973) regressions ranked by he age of he firms since heir IPO. Using price-value raio as a direc measure of marke misvaluaion, I find ha firm end o reduce is leverage when is equiy is overvalued and more imporanly, his change in leverage is mainly driven by equiy issues, no change in reained earnings or deb issues. I also find ha here is a srong negaive relaionship beween firm s curren leverage and he iming measure of hisorical misvaluaion. Low leverage firms are hose ha raised funds when heir 2 Among hese, Frankel and Lee (1998) find ha he residual income model has predicive power in he cross-secion of sock reurns in domesic and inernaional markes. Penman and Sougiannis (1998) also find suppor for he valuaions performance of he model. 4

10 marke valuaions were high in he pas hisory. This is consisen wih he heory of equiy marke iming. In addiion, I also analyse he long-erm influence of marke iming on capial srucure. I find ha he marke iming effec on firms accumulae change in leverage since IPO is negaive and significan. Firms do no end o adjus heir leverages afer IPO. This is inconsisen wih he dynamic models of rade-off heory wih adjusmen cos. My resuls also show ha he curren marke iming measure sill inversely affec he capial srucure even afer 10 years, suggesing ha he influence of pas marke misvaluaion on capial srucure is quie persisen. The resul is robus o boh leverage raios defined in book and marke values, and i also holds regardless wheher ex-pos earnings or analys forecass are used o compue he inrinsic value. Takes ogeher, he resuls indicae ha marke iming plays an imporan role in shaping financing aciviy boh in he shor-erm and in he long-run. Using a direc misvaluaion measure, my resuls provide he suppor for he view of Baker and Wurgler (2002) ha capial srucure is he cumulaive oucome of pas aemps o ime he equiy marke. The res of he paper is organized as follows. Secion 2 reviews prior lieraure on capial srucure. Secion 3 describes he daa and consrucs he iming measure used in his paper. Secion 4 describes he empirical mehodology o be adoped wihin his paper. Secion 5 presens he empirical resuls. Secion 6 ess he robusness of he resuls and Secion 7 concludes. 5

11 2. Exising Capial Srucure Research In perfec and efficien markes wih no fricions such as ransacion coss and axes, Modigliani and Miller (1958) show ha he value of he company is independen of is capial srucure. This forms he basis for he subsequen research on capial srucure. I also provides he base on which one can examine he real world reasons why capial srucure is relevan. By relaxing he assumpions of perfecions and efficiencies, various heories have been developed hrough ime. Trade-off Theory By adding various imperfecions, including axes, coss of financial disress, and agency coss, bu reaining he assumpions of marke efficiency and symmeric informaion, he rade-off heory focuses on he rade-off beween he cos and benefi of deb. When he firm finances wih deb, here is ax benefi from he ineres ax shield. However, a he same ime, he cos of deb arises, such as he cos of financial disress including bankrupcy cos of deb and non-bankrupcy cos. The rade-off heory predics ha here exiss a poin in a firm s capial srucure ha maximizes he firm value, a which he marginal benefi of using deb is jus offse by he marginal cos of i. This rade-off heory quickly ranslaes o he empirical hypoheses. For example, i predics reversion of he acual deb raio owards a arge or opimum, and i predics differen arge leverage raios for firms in differen indusries. Evidence supporing he rade-off heory is mixed. Schwarz and Aronson (1967) find evidence of srong indusry effecs in deb raios, which hey inerpre as evidence 6

12 of opimal raios. Mackie-Mason (1990) repors evidence ha firms wih ax loss carry forwards are less likely o issue deb. This is consisen wih Miller and Modigliani (1966). They find he posiive effecs of ineres ax shields in he marke values of elecric uiliies. There is more direc evidence ha firms adjus oward a arge raio. Marsh (1982), Hovakimian, Opler, and Timan (2001), Korajczyk and Levy (2003), Hovakimian (2004) confirm he role of he arge leverage in share issuance and buyback. Frank and Goyal (2004) examine he relaive imporance of 39 facors in leverage decisions, and argue in favour of he rade-off heory. Leary and Robers (2005) show ha firms are inacive wih respec o heir financial policies, bu do issue or buy back shares in he way o adjus oward arge leverage. However, Timan and Wessels (1988), Rajan and Zingales (1995), Fama and French (2002) and many oher sudies find ha negaive relaionship beween firm s profiabiliy and heir leverage raio. This is inconsisen wih he rade-off predicion ha more profiable firms would borrow more o reduce heir ax liabiliies. Jalilvand and Haris (1984) and Fama and French (2002) find ha he speed of adjusmen oward arge leverage is very slow. There are recen sudies quesioning he exisence of a arge leverage raio. Baker and Wurgler (2002) show ha a firm s curren capial srucure is he cumulaive resul of pas hisorical marke-o-book raio and he effec is permanen. Welch (2004) finds ha prior sock reurns are he main deerminan of change in marke leverage, and firms do no acively offse he effecs of sock reurns on heir capial srucure. Chang and Dasgupa (2006) argue ha he rebalancing can be obained even firms make heir deb or equiy issuance decisions randomly and he leverage mean reversion is merely a mechanical effec. 7

13 Pecking Order Theory Pecking order heory is formally developed by Myers (1984) and Myers and Majluf (1984). Due o adverse selecion, ouside invesors are willing o buy securiies a a discoun. In his case, firms prefer inernal o exernal finance. In pecking order heory, here is no opimal capial srucure. When ouside funds are needed, firms prefer deb o equiy because of lower informaion coss associaed wih deb issues. Exernal equiy is only used as a las resor. The informaion asymmery is no he only reason for pecking order. Donaldson (1961) finds ha ransacion coss can be he reason for he prioriy of he source of financing. Recen papers by Lee (1997), Heaon (2002), and Hackbarh (2003) argue ha managerial opimism can be he driven force for he pecking order. Opimisic managers are unwilling o issue equiy because hey hink heir socks are undervalued. Several papers have been able o find ha here are insances where pecking order heory is a good approximaion o realiy. Shyam-Sunder and Myers (1999) es he hypohesis ha he financing defici should be mached dollar-for-dollar by a change in corporae deb, which is a predicion of pecking order heory. Using a sample of 157 firms ha raded coninuously from 1971 o 1989, he simple pecking order model ouperforms he arge adjusmen model in explaining he ime series variaion in deb raio. Pecking order heory also helps o explain he phenomenon ha more profiable firms would like o issue less deb due o inernal generaed fund is preferred. Fama and French (2002) find ha more profiable firms are less levered, consisen wih he predicions of pecking order. 8

14 Bu he ess of he pecking order heory have no been able o show ha i is of firs order imporance in deermining firm s capial srucure. Helwege and Liang (1996) find he exernal financing is unrelaed o he shorfall in inernally generaed funds for firms wen public in 1983, inconsisen wih he pecking order heory. Chirinko and Singha (2000) criicize he validiy of pecking order ess of Shyam-Sunder and Myers. They show ha he eleganly simple es generaes misleading inferences when evaluaing plausible paerns of exernal financing. Frank and Goyal (2003) exend he pecking order es for a much larger sample of US firms. Their resuls are conrary o he pecking order heory. They find ha ne equiy issues rack he financing defici more closely han do ne deb issues. Afer including convenional leverage facors, financing defici is less imporan in explaining ne deb issues over ime for firms of all sizes. Fama and French (2005) challenge he pecking order heory by showing ha firms frequenly issue and repurchase equiy. Marke Timing Theory By relaxing he assumpion of marke efficiency, marke iming heory predics ha he managers aemp o ime equiy markes by issuing shares a high marke prices and repurchasing shares a low marke prices. The inenion is o exploi emporary flucuaions in he cos of equiy relaive o he cos of oher forms of capial. Because he marke iming heory does no rely on he assumpion of semi-srong marke efficiency, miss-pricing can occur. Managers know more informaion abou heir firms, hey will ake he opporuniies by issuing or buying back shares, evenually change he firms capial srucures. 9

15 In pracice, corporae execuives seem o acively engage in marke iming in heir financing decisions. A survey conduced by Granham and Harvey (2001) find ha wo-hirds of CFOs agree ha he amoun by which our sock is undervalued or overvalued was an imporan or very imporan consideraion in issuing equiy. In ha survey as a whole, equiy marke prices are regarded as more imporan han 9 ou of 10 oher facors considered in he decision o issue common sock, and more imporan han all 4 oher facors considered in he decision o issue converible deb. The marke iming heory is only developed recenly and i has generaed significan conroversy because i is a odds wih he radiional heories of capial srucure. The empirical evidence is mixed. Baker and Wurgler (2002) propose his alernaive heory explaining firms capial srucure. In heir empirical sudy, Baker and Wurgler consruc an exernal finance weighed hisorical marke-o-book raio, EFWAMB, o capure firms pas equiy marke iming aemps. They find ha, conrolling for firms growh opporuniies using curren marke-o-book raio, leverage is negaively relaed o EFWAMB, which hey inerpre as evidence supporing he equiy marke iming heory. They also find ha he resuling effecs on capial srucure are very persisen and conclude he capial srucure is he cumulaive oucome of pas aemps o ime he equiy marke. 1 M es ds M efwa, 1 1 B s 0 B er dr r 0 ( ) ( ) (1) s However, here are wo broad criicisms having been levelled a Baker and Wurgler s findings. Several recen sudies quesion he persisen impac of marke 10

16 iming aemps. Leary and Robers (2005) show ha a ypical US firm rebalances is capial srucure in hree o five years following he equiy issuance. Similarly, Flannery and Rangan (2006), Kayhan and Timan (2007), Ali (2006) and Hovakimian (2006) sugges ha he impac of marke iming on leverage vanishes very quickly and no evidence of persisence being found. However, Huang and Rier (2007), using aggregae measures of marke valuaion, find evidence for a long lasing marke iming effec on capial srucure. The second criicism is on EFWAMB ha Baker and Wurgler used for equiy marke iming measure. Hovakimian (2006) argues ha he negaive relaionship beween marke-o-book and leverage is no indicaive of marke iming, bu a proxy for firms growh opporuniies. Ali (2006) isolaes he iming aemps in a single major financing even, he iniial public offering, and find ha he impac of equiy marke iming on leverage is shor-lived. In recen work, Kayhan and Timan (2007) poin ou ha he significance of he hisorical marke-o-book may be due o he noise in he curren marke-o-book raio. Ellio, Koëer-Kan and Warr (2007) es he simple pecking order model using direc value measure and heir resuls provide evidence suppor for Baker and Wurgler. Based on he resuls from US daa, he validiy of he marke iming hypohesis remains unresolved. There are some evidence findings of equiy marke iming by managers in counries oher han US. Hogfeld and Oberonko (2004) and Bie and Haan (2004) find inverse relaion beween EFWAMB and leverage in Sweden and Neherlands, respecively. In conras, Mendes, Kayo and Basso (2005) sudy he issue in Brazil and 11

17 find no relaionship beween EFWAMB and leverage. Similarly, Mahajan and Tararoglu (2007) invesigae he marke iming hypohesis in G-7 counries oher han US. Excep Japan, he firms in all he oher counries undo he effec of equiy issuance and he impac of equiy marke iming aemps on leverage is shor lived. Their resuls are inconsisen wih he predicion of he equiy marke iming heory and more in line wih dynamic rade-off model. 3. Daa and Marke Timing Measure 3.1. Daa and Conrol Variables My iniial sample comprises all firms on COMPUSTAT during he period The regulaed firms (SIC codes beween 4900 and 4999) and financial firms (SIC codes beween 6000 and 6999) are no included. 4 A small number of firms wih invalid COMPUSTAT forma codes 5 and firms wih book value of asses less han $10 million or wih missing values of relevan variables are also excluded from he sample. Sock price and reurn daa are obained from he Cener for Research in Securiies Prices (CRSP), while I/B/E/S is used for analys forecass in he residual income model. I define book deb as oal asses (COMPUSTAT Annual Iem 6) minus book equiy. Book equiy is defined as oal asses less oal liabiliies (Iem 181) and preferred sock (Iem 10) plus deferred axes (Iem 35). When preferred sock is missing, i is replaced wih he redempion value of preferred sock (Iem 56). Hovakimian (2006) poins ou ha Baker and Wurgler s reamen of converible deb is 3 Availabiliy of analys forecass from I/B/E/S limis my sample period o firms laer han The capial srucures for hese firms are eiher regulaed or highly leveraged. 5 These are firms which have value of 4, 5 or 6 for he daa iem 318 in COMPUSTAT. 12

18 no consisen. They only include converible deb in he book value of equiy bu no in he marke value of equiy. Following mos of previous sudies of capial srucure, I rea converible deb as deb raher han as equiy. Marke equiy is defined as common shares ousanding (Iem 25) imes fiscal year closing price (Iem 199). Then marke value of asses is he resul of oal asses minus book equiy plus marke equiy. I define leverage in wo ways based on book and marke value measures. Book leverage is defined as book deb divided by oal asses. Marke leverage is defined as book deb divided by marke value of asses. I drop firm-year observaions where he book leverage and marke leverage exceeds 100%. Ne equiy issues, e/a, is he change in book equiy minus he change in reained earnings. I define newly reained earnings, RE/A, as he change in reained earnings (Iem 36). I define ne deb issues, d/a, as he change in book deb. All hese hree variables are normalized by fiscal year-end oal asses and expressed in percenage erms. Conrol variables include 5 variables from Fama and French (2002). Profiabiliy is measured by EBITDA/A, which is earnings before ineres, axes, and depreciaion (Iem 13) divided by oal asses and expressed in percenage erms. Size is measured as he naural logarihm of ne sales (Iem 12) in millions of 2000 dollars. Asse angibiliy, PPE/A, is defined as ne plan, propery, and equipmen (Iem 8) scaled by oal asses and expressed in percenage erms. R&D/A is he percenage erm of he research and developmen expense (Iem 46, replaced by zero when missing) divided by oal asses. RDD is a dummy variable ha akes he value of one when Iem 13

19 46 is missing. Marke-o-book raio is marke value of asses on oal asses. Following Baker and Wurgler (2002), I drop observaions where he marke-o-book raio is above 10. To minimize he influence of ouliers, I winsorize he sample a 0.5 and 99.5 perceniles Residual Income Model using Analyss Forecass The Residual Income Model or someimes known as Edwards-Bell-Ohlson (EBO) valuaion model calculaes he fundamenal value of he firm by anchoring is price a he curren book value of equiy and adding a premium o book value based on fuure residual earnings. This echnique has been used exensively o calculae fundamenal value in previous sudies. In addiion, Felham and Ohlson (1995) show ha he model is equivalen o he heoreically sound dividend discoun model (DDM) under clean surplus accouning. 6 Following Lee, Myers and Swaminahan (1999), I use he wo-sage model o esimae he inrinsic value: (1) forecas earnings explicily for he nex hree years, and (2) afer hree years, assumes ha firms grow sably a a rae equal o median indusry ROE. The inrinsic value is calculaed by he equaions as follows: ( FROE r ) ( FROE r ) E( V ) B B B TV (2) 1 e 2 e 2 1 (1 re) (1 re) TV ( FROE re) B (1 r ) r 3 2 e e 2 (3) where 6 Clean-surplus accouning describes he siuaion where he change in book value is equal o earnings minus dividends for he given period. 14

20 B = book value from he mos recen financial saemen divided by he number of shares ousanding in he curren monh from I/B/E/S. re = he cos of equiy FROE+i = forecased ROE for period +i. For he firs hree years, his variable is compued as FEPS+i/B+i-1, where FEPS+i is he I/B/E/S mean forecased EPS for year +i and B+i-1 is he book value per share for year +i-1. 3 years of forecas period may appear o be shor for capuring he fuure growh opporuniies of a firm. However, Lee, Myers and Swaminahan (1999) repor ha he resuls wih 3-year forecas horizon are very similar o hose obained using 12-year forecas horizon. In fac, he residual income model does no capure raw earnings. Raher i employs abnormal earnings which end o rever o zero. By aking year 3 abnormal earnings as perpeuiy, I implicily assume mean reversion of he abnormal earnings over ime. I use Fama and French s (1997) hree facor model o calculae he cos of equiy capial. This model segregaes firms ino 48 indusry classificaions and creaes replicaing porfolios based on size and book-o-marke characerisics. Explicily, he risk premiums repored for each indusry on a monhly basis are combined wih he effecive annual risk-free rae, based on he curren monhly risk-free rae, o generae an indusry cos of equiy for ha given monh. 7 The shor-erm T-bill rae is used here for he risk-free ineres rae. Lee, Myers and Swaminahan (1999) find ha by using 7 Cos of equiy = (Fama-French risk premium + 12 monhly risk-free rae) /

21 shor-erm ineres raes raher han long-erm reasury bonds raes, heir value-price raio have a much lower sandard deviaion and a faser rae of mean reversion. One-year-ahead (FEPS+1) and wo-year-ahead (FEPS+2) EPS forecass are colleced from I/B/E/S from I/B/E/S also repors esimaes of he long-erm growh rae (Lg) ha can be used o compue he hree-year-ahead EPS: FEPS+3=FEPS+2 (1+Lg). 8 I esimae he dividend payou raio by dividing acual dividends from he las fiscal year (Iem 21) by earnings over he same ime period (Iem 237). Following Lee, Myers and Swaminahan (1999), for firms experiencing negaive earnings, he payou raio is esimaed by dividing he dividend paid o 0.06 imes oal asses. Payou raios of less han zero or greaer han one are assigned o zero or one. Under he clean-surplus accouning, I can forecas he fuure book values per share and ROE using hese earning forecass and he dividend payou raio. The misvaluaion is hen calculaed as he raio of marke price of he sock over he inrinsic value of he sock from he residual income model. PV P EV ( ) (4) If here is no misvaluaion, PV should be equal o one. If PV is greaer (lower) han one, i implies overvaluaion (undervaluaion). I eliminae he firm-year observaions where is PV exceeds The brief summary saisics of PV for he pooled sample is shown in Table 1. 8 Prior o 1981, I/B/E/S does no repor Lg. When his variable is missing, I use he composie growh rae implici in FY1 and FY2 o forecas FY3. 9 The resuls (no abulaed) are qualiaively very similar when he observaions wih PV greaer han 10 are kep in he sample. 16

22 3.3. Exernal Finance Weighed Average Price-Value Raio Baker and Wurgler define exernal finance weighed average marke-o-book raio (EFWAMB) as he proxy for equiy marke iming. In heir model, he misvaluaion is capured by he marke-o-book raio. However, in my model, I am able o esimae he misvaluaion direcly hrough price-value raio. Then he iming proxy developed from his direc measure of misvaluaion should capure he marke iming more precisely. Following he inuiion of Baker and Wurgler s EFWAMB, my marke iming proxy is defined as 1 P es ds P efwa, 1 1 s V s 0 V er dr r 0 ( ) ( ) (5) I calculae EFWAPV saring from he firs year (r=1 and s=1) ha he price-value raio is available. e and d denoe ne equiy issue and ne deb issue defined in he previous secion. Following Baker and Wurgler (2002), for he purpose of compuing his variable, I se he minimum weigh o zero. The negaive weighs are also se o zero o ensure ha I am forming a weighed average. A zero weigh jus means ha here is no informaion conained in he price-value raio in ha paricular year. The maximum EFWAPV is se o en in order o limi he effec of ouliers. The summary saisics of his variable is conained in Table 1. This variable akes high values for firms ha raised exernal finance when he price-value raio was high and vice-versa. More weighs are given o valuaions ha prevailed when significan exernal financing decisions were made; no maer i is deb 17

23 issue or equiy issue. I is beer han jus using lagged price-value raio as i capures he pracical opporuniies for marke iming ha firms have been aking Summary Saisics Table 1 repors he summary saisics for he pooled samples. There are 41,458 firm-year observaions from 1976 o Some ineresing saisics are as follows. The median of R&D/A is zero, suggesing ha a leas half of he firms in he sample do no have informaion abou heir research and developmen expense. Therefore, i is imporan o include a dummy variable o conrol for firms wih no research and developmen expense. Average P/V for he pooled sample is 1.36, implying ha firm in he sample is slighly overpriced on average. The saisics for oher conrol variables are very similar o previous capial srucure sudies. 10 [ inser Table I here ] 4. Empirical Mehodology This secion presens my empirical mehodology. Firs, I discuss he empirical model used for esing he marke iming effec on annual change in leverage. Then, I presen he model esing he relaionship beween hisorical marke valuaion and firm s leverage. This is followed by discussing a series of regressions implemened o examine he persisence of marke iming effec. Finally, his secion is concluded wih a brief summary on Fama-MacBeh regression mehod used in he paper. 10 Baker and Wurgler (2002), Liu (2007), Mahajan and Tararoglu (2007). 18

24 4.1. Deerminans of Annual Changes in Leverage Marke iming heory suggess ha over-valued firms end o issue shares o decrease heir leverage raios, implying negaive relaionship beween he measure of misvaluaion and annual changes in leverage. However, he negaive relaionship is only parial evidence for marke iming heory as firms may repay heir deb o decrease leverage when heir sock is over-priced. In ha case, here is no aemp of marke iming. Therefore, I hen decompose he change in leverage o examine wheher he effec comes hrough ne equiy issuance, as marke iming implies. The esing equaion on annual changes in leverage is: D D P M EBITDA PPE R & D D ( ) ( ) 1 a b( ) 1 c( ) 1 d( ) 1 esize 1 f ( ) 1 g( ) 1 hrdd 1 i( ) 1 (6) A A V B A A A A and where he dependen variable is he annual changes in leverage. On he lef hand side, he price-value raio consruced in he previous secion is used here as a proxy for misvaluaion. Oher conrol variables described in he daa secion are also included here. They are marke-o-book (conrolling for invesmen opporuniies), firm profiabiliy, size, asse angibiliy, research and developmen expense, and a dummy variable ha akes he value of one for firms reporing no research and developmen. In paricular, he lagged leverage is also included in he regressions as leverage is bounded beween zero and one. Leverage can change only in one direcion when i is near one end of boundaries regardless of he oher variables value. No conrolling for lagged leverage may obscure he effecs of he oher variables. Lagged leverage herefore should ener wih a negaive sign. All conrol variables are a one year lag. If sock 19

25 misvaluaion has any effec on change in leverage as marke iming heory implies, he coefficien for misvaluaion proxy should be negaive. To examine he marke iming effec on annual changes in leverage associaed wih ne equiy issuance, I decompose he change in leverage as follows: D D E E ( ) ( ) 1 [( ) ( ) 1] A A A A e RE 1 1 = ( ) ( ) [ E 1( A A A A )] (7) 1 The firs erm on he righ-hand side is he negaive of he ne equiy issues on oal asses in year. If all of he new equiy capial is used o pay down deb, hen he change in leverage resuling from equiy issues is equal o he negaive of he amoun issued. However, o he exen ha he new equiy capial adds o oal asses, he reducion in leverage is less han one-for-one. The second erm is he change in reained earnings. Newly reained earnings add o equiy capial and hence decrease leverage. The final erm is he residual change in leverage, which depends on he oal asses growh from he combinaion of equiy issues, deb issues, and newly reained earnings. I regress each of hese hree componens of annual changes in leverage on he same conrol variables used in Equaion (6). By doing so, i allows me o deermine wheher marke iming affecs leverage hrough ne equiy issues, as suggesed by marke iming heory Deerminans of Leverage If firms have no arge leverage raio, he effecs of marke iming will accumulae and finally drive he firms capial srucure. To examine he effec of 20

26 hisorical marke valuaion on firm s curren srucure, I use he regression equaion as follow: D P M EBITDA PPE R & D ( ) a b( ) efwa, 1 c( ) 1 d( ) 1 esize 1 f ( ) 1 g( ) 1 hrdd 1 (8) A V B A A A where he leverage is measured in boh book and marke value. P/Vefwa consruced before is included here as a conrol variable. I capures he effec of firm s hisorical misvaluaion on is capial srucure. If marke iming heory is rue, he firm will ake he opporuniies o issue equiy as long as is sock is over-priced by he marke, implying ha he accumulae effec of marke iming on capial srucure is negaive. The oher conrol variables included in he regression are marke-o-book raio, profiabiliy, size, angibiliy, research and developmen expense, and a dummy variable ha akes he value of one for firms reporing no research and developmen. All conrol variables are a one year lag Persisence The marke iming heory becomes a lo conroversial in academic world when he longer ime horizon is considered. Baker and Wurgler (2002) find ha hisorical marke valuaions have large and very persisen effecs on firms capial srucure. Huang and Rier (2007) use aggregae measures of marke valuaion, find evidence of a long lasing marke iming effec. On he oher hand, Ali (2006) uses a ho marke dummy in his sudy and conclude ha he long-run impacs of marke iming on leverage are limied. Resuls from Kayhan and Timan (2007) indicae ha alhough firms hisories srongly influence heir capial srucures, over ime he capial srucure end o move oward o arge leverage raio. Therefore, answering he quesion ha 21

27 how persisen is he impac of marke iming on capial srucure has become very vial here. If he persisence is rejeced, all he previous resuls migh jus be he evidence supporing he dynamic rade-off heory ha he firms adjus heir capial srucure around a arge rae hrough ime, which suggess marke ime is no of firs-order imporance. In his par of secion, a series of regressions are implemened o es he degree and magniude of marke iming persisence, including deerminans of cumulaive changes in leverage since IPO and he effec of lagged marke iming measure on he capial srucure. IPO iself usually is a grea opporuniy for marke iming, I firs examine wheher his iming effec is reversed in he years subsequen o he IPO. The es regression is as follows: D D P M EBITDA PPE R & D D ( ) ( ) IPO a b( ) efwa, 1 c( ) 1 d( ) 1 esize 1 f ( ) 1 g( ) 1 hrdd 1 i( ) IPO (9) A A V B A A A A The dependen variable is he accumulaive change in leverage since IPO-year. The conrol variables are he ones used in Equaion (8), excep ha he leverage from is IPO level is also included here. The reason of his is he same as including he lagged leverage in Equaion (6). If marke iming has a permanen effec on leverage, hen he cumulaive change in leverage from is IPO level should coninue o reflec he marke iming effec in he years afer he IPO, implying he negaive sign for he marke iming coefficien. I hen perform some direc ess on he persisence of he pas marke iming on firm s curren capial srucure. 22

28 D P M EBITDA PPE R & D ( ) a1 b1 ( ) efwa, 5 c1( ) 5 d1( ) 5 e1size 5 f1( ) 5 g1( ) 5 h1rdd 5 (10) A V B A A A D P M EBITDA PPE R & D ( ) a2 b2 ( ) efwa, 5 c2( ) 1 d2( ) 1 e2size 1 f2( ) 1 g2( ) 1 h2rdd 1 (11) A V B A A A The firs regression looks a wheher he iming a -5 has any impac on curren leverage, conrolling for oher firm characerisics a -5. The second regression looks a he effec of he iming a -5 on curren capial srucure wih mos up-o-dae value of conrol variables. I expec he signs for boh iming measure coefficiens o be negaive if marke iming has persisen effec. 5 year may be no long enough in he academic sudies. This persisence es is repeaed wih 10 years lag o find ou wheher he marke iming a -10 has any impac on curren capial srucure. D P M EBITDA PPE R & D ( ) a1 b1 ( ) efwa, 10 c1( ) 10 d1( ) 10 e1size 10 f1( ) 10 g1( ) 10 h1rdd 10 (12) A V B A A A D P M EBITDA PPE R & D ( ) a2 b2 ( ) efwa, 10 c2( ) 1 d2( ) 1 e2size 1 f2( ) 1 g2( ) 1 h2rdd 1 (13) A V B A A A Again, persisence effec predics ha he pas marke iming measure is negaively associaed wih curren leverage in boh cases Fama-MacBeh Regression Mehod All he ess are esimaed using Fama-MacBeh regressions. I is a panel sudy where he ime variable is defined by he age of he firms since heir IPO. I firs esimaes he regressions for sub-samples wih differen ages, and hen hese esimaion resuls are averaged o repor he resuls for he whole sample. I is appropriae here due o he high degree of dependence beween successive observaions of he iming 23

29 variables. I also esimae hese regressions using OLS wih indusry fixed effecs defined by SIC codes and obain similar resuls Empirical Resuls 5.1. Deerminans of Annual Changes in Leverage The analysis begins by firs esablishing wheher firm end o ime equiy marke when is sock is over-priced. Column 1 of Table II repors he regression esimaes for Equaion (6). The prior year misvaluaion is found o have negaive effec on annual changes in leverage a 5% significance level. One sandard deviaion (1.02) increase in price-value raio is associaed wih ( =0.173) decrease in leverage. This is consisen wih he idea ha firms increase equiy when heir socks are overvalued by marke. However, i does no rule ou he possibiliy ha lower leverage comes from higher reained earnings or lower deb. The res of Table II repors he regression resuls for he decomposiion componens, from which i allows me o deermine wheher marke iming affecs leverage hrough ne equiy issues, as suggesed by marke iming heory. By he decomposiion Equaion (7), he sum of he coefficiens of each independen variable for he hree decomposiion erms should be equal o he oal effec of changes in leverage. Column 2 shows ha price-value raio has a posiive relaion wih ne equiy issues and he coefficien is significan a 1% level. 12 One sandard deviaion (1.02) increase in price-value raio is associaed wih ( =0.704) increase in ne equiy issues. Column 3 shows ha overpriced firm end o reduce is reained earnings 11 The OLS resuls are no repored here o save space. 12 I has a negaive coefficien and he dependen variable is negaive of ne equiy issues. 24

30 and increase leverage. The las column predics ha price-value raio is posiively relaed o growh in asses, an effec ha ends o increase leverage. Therefore, negaive effec of change in leverage is mainly from ne equiy issues hough he increase of ne equiy issues does no reduce leverage one for one because of increasing leverage effec from oher componen. The coefficiens for oher conrol variables are also repored. I shows ha marke-o-book, profiabiliy and R&D expenses end o reduce leverage, while size and angibiliy end o increase leverage. The esimaes of hese coefficiens are consisen wih prior capial srucure sudies. [ inser Table II here ] 5.2. Deerminans of Leverage Table III repors he cumulaive effec of pas marke iming on leverage. The esimaed P/Vefwa,-1 coefficiens for book leverage and marke leverage are and respecively and boh significan a 1% level. In erms of economic magniudes, one sandard deviaion increase in P/Vefwa is associaed wih a decrease in book leverage of 2.09 percenage poins and a decrease in marke leverage of 3.33 percenage poins. 13 These resuls sugges ha hisorical misvaluaion of firm s value is imporan in explaining he curren level of leverage, and here is a negaive accumulaed effec from pas marke iming on leverage raio. 13 Sandard deviaion of P/V efwa is

31 Marke-o-book which is conrolling for invesmen opporuniies ends o reduce he leverage raio, especially on marke leverage. More profiable firms are likely o have lower leverage raio. Size and angibiliy end o increase leverage raio on average, bu he effec of angibiliy on book leverage is no significan from zero. I also find ha here is asymmeric effec from R&D expense on capial srucures. [ inser Table III here ] 5.3. Persisence In his par of secion, I es he persisence of marke iming begin wih sudying wheher firm adjuss is capial srucure afer IPO-year. Table IV repors he regression resuls for he esing Equaion (9). The marke iming has a negaive relaion wih accumulae change in book leverage since IPO and he coefficien is significan a he 5% level. The iming effec is more pronounced when he accumulae change measured in marke leverage, and i is significan a he 1% level. This persisence of marke iming effec on capial srucure conrass wih he rade-off heory in which firm adjuss is leverage back o is arge raio afer IPO. Ineresingly, he coefficiens -saisics for he oher conrol variables are very similar o he resuls from Ali (2006). However, his ho marke dummy does no capure he persisence effec of marke iming on he accumulae change in leverage. Mos of his resuls are no significanly differen from zero even a he 5% level. His iming proxy is defined based on he monhly IPO volume. I is raher an aggregae measure of ho equiy marke han a misvaluaion measure specific o he individual 26

32 firms. Marke iming heory is based on he assumpion of misvaluaion for a paricular firm s equiy value. Even in a ho marke, here are some firms which are undervalued. Similarly, overvalued firms also exis in a cold marke. Comparing o Ali s marke iming measure, my P/Vefwa is more appropriae o capure cross-secional difference in misvaluaion and marke iming. This migh be he reason why ha I find persisence iming effec on he accumulae changes in leverage bu Ali canno. [ inser Table IV here ] I hen examine he effec of 5-year-lagged marke iming measure on curren leverage raio. The firs wo columns of Table V repor he resuls for wo regressions when book leverage raio is used. The coefficien of P/Vefwa,-5 is and significanly differen from zero a he 1% level, conrolling for firm characerisics a -5. Reading from he second regression resuls, he coefficien of P/Vefwa,-5 is The effec of marke iming is even greaer when I conrol for firms oher characerisics a 1 year lag. The hird and fourh columns of Table V presen he regression resuls for marke leverage. When marke leverage is used, he effec of he marke iming is more pronounced in boh regressions. Resuls in column 2 and 4 of Table V are very similar o hose in Table II. In fac, he only difference here is ha I replace P/Vefwa,-1 wih P/Vefwa,-5. However, he coefficien for my iming proxy is very similar, implying ha he marke iming effec is highly persisen in inermediae erm. 27

33 [ inser Table V here ] I finish his secion by exending he persisence es furher o 10 years. All he regression resuls are presened in Table VI. Because a much longer ime horizon is used here, he number of firm-year observaions drop o 9691 for book leverage and 9670 for marke leverage. However, he marke iming effecs on capial srucure are srong and negaive for boh book leverage and marke leverage. Comparing o Table V, he coefficiens for he iming proxy are close enough ha I can conclude he long-erm effec of marke iming. My resuls do no suppor he dynamic rade-off heory as he P/Vefwa remains a srong deerminan of capial srucure even afer 10 years. The resuls are consisen wih Baker and Wurgler s finding ha firm s curren capial srucure is he cumulaive oucome of pas aemps o ime he equiy marke. [ inser Table VI here ] 6. Robusness Check The above menioned resuls are based on I/B/E/S earning forecass used in he residual income model. Because I/B/E/S hisorical earnings forecass are only available for US firms afer 1976, i places some limiaions on my sample size. On he oher hand, mainly large firms are followed by analyss, hus he sample can be biased in favour of large maure firms. To es he robusness of my resuls, I re-esimae he price-value raio using he perfec forecas version of he residual income model. 14 In his implemenaion, he ex-pos realizaion of earnings is used insead of analys 14 More deail see D Mello and Shroff (2000). 28

34 forecas. I increases he number of observaions, bu also imposes a minimum fouryear survival-ship bias on he firms in my sample because of he valuaion model I used. 15 Table VII repors he resuls for he deerminans of leverage. The resuls are very similar o hose in Table II wih he sample size doubled here. Bu he resuls are essenially he same. I sill has a srong negaive impac on he leverage raio. The coefficiens of lagged marke iming measure are significan a he 1% level in boh cases. All oher conrol variables are consisen wih my previous resuls. [ inser Table VII here ] I hen examine he persisence of marke iming effec on leverage. Table VIII shows he resuls for he wo regressions esing persisence. Comparing his wih he resuls using analys forecass, he number of he firm-year observaions increase o for book leverage and for marke leverage. However, all he coefficien esimaes are consisen wih previous analysis. Even perfec forecas version of he residual income model is implemened in calculaing he iming measure, he effec of marke iming on leverage is sill highly persisen no maer which measure of leverage raio is used. [ inser Table VIII here ] 15 In his secion, he sample comprises firms on COMPUSTAT from 1970 o

35 7. Conclusion The sudy of equiy marke iming effec on capial srucure has generaed ho debae in he lieraure. Bu he commonly used measures of marke iming are ofen oo noisy o capure he exen of marke iming from equiy misvaluaion. The evidence is sill inconclusive especially when i comes o he long-erm effec of marke iming. In his paper, I use he residual income model o direcly measure he fundamenal value of a firm s equiy. The use of his mehod gives me a clean measure of mispricing and allows me o disenangle he muliple inerpreaions associaed wih marke-o-book raio. I hen develop he exernal finance weighed average price-value raio as he direc measure of marke iming from equiy misvaluaion. Using his more appropriae iming measure, I find ha over-valued firm is more likely o issue equiy o reduce is leverage. My findings are robus o oher facors ha have been found o significanly impac he firm s capial srucure. This negaive relaionship is apparen wheher leverage is measured in book or marke value. In addiion, I find ha he marke iming effec on capial srucure can las for a leas 10 years. This canno be explained by he dynamic rade-off heory wih adjusmen cos. To summarize, he resuls in his paper suppor he hypohesis ha firm s equiy misvaluaion has significan long lasing effecs on is capial srucure. I is consisen wih he marke iming heory suggesed by Baker and Wurgler (2002). 30

36 Reference Ali, A., 2006, How persisen is he impac of marke iming on capial srucure? Journal of Finance 61, Baker, M. and J. Wurgler, 2002, Marke iming and capial srucure, Journal of Finance 57, Bie, T. and L. Haan, 2004, Does marke iming drive capial srucures? A panel daa sudy of Duch firms, Working paper, De Nederlandsche Bank. Chang, X. and S. Dasgupa, 2006, Targe behaviour and financing: How conclusive is he evidence? Working paper, Universiy of Melbourne and Hong Kong Universiy of Science & Technology. Chirinko, R.S. and A.R. Singha, 2000, Tesing saic radeoff agains pecking order models of capial srucure: A criical commen, Journal of Financial Economics 58, D'Mello, R. and P.K. Shroff, 2000, Equiy undervaluaion and decisions relaed o repurchase ender offers: an empirical invesigaion, Journal of Finance 55, Donaldson, G., 1961, Corporae deb capaciy: A sudy of corporae deb policy and he deerminaion of corporae deb capaciy, Harvard Business School, Division of Research. Ellio, W.B., J. Koëer-Kan and R.S. Warr, 2007, A valuaion-based es of marke iming, Journal of Corporae Finance 13, Fama, E. and K. French, 1997, Indusry coss of equiy, Journal of Financial Economics 43, Fama, E. and K. French, 2002, Tesing radeoff and pecking order predicions abou dividends and deb, Review of Financial Sudies 15, Fama, E. and K. French, 2005, Financing decisions: Who issues sock? Journal of Financial Economics 76, Felham, G.A. and J.A. Ohlson, 1995, Valuaion and clean surplus accouning for operaing and financial aciviies, Conemporary Accouning Research 11, Flannery, M.J. and K. Rangan, 2006, Parial adjusmen oward arge capial srucures, Journal of Financial Economics 79, Frank, M. and V. Goyal, 2003, Tesing he pecking order heory of capial srucure, Journal of Financial Economics 67,