The Impact of Ownership Structure on Debt Financing of Japanese Firms With the agency cost of Free Cash Flow

Transcription

1 The Impac of Ownership Srucure on Deb Financing of Japanese Firms Wih he agency cos of Free Cash Flow Lingling Wu* Deparmen of Economics and Finance Ciy Universiy of Hong Kong January 12, 2004 * Conac informaion: Lingling Wu, Deparmen of Economics and Finance, Ciy Universiy of Hong Kong, Ta Chee Avenue, Kowloon, Hong Kong @suden.ciyu.edu.hk. Phone: (852) Fax: (852)

2 The Impac of Ownership Srucure on Deb Financing of Japanese Firms Wih he agency cos of Free Cash Flow Absrac This paper, using 833 observaions of lised Japanese firms beween he years 1992 o 2000, explore he implicaions of he free cash flow hypohesis concerning he disciplinary role of ownership srucure in corporae capial srucure policy. The resuls show (1) The relaion beween leverage and free cash flow are significanly posiive and greaer for low-growh firms han for he high-growh firms. (2) In he firms wih low growh opporuniies, insiuional invesors discourage managerial overspending by governance process and hence compensae for he deb monioring. The managerial ownership may align he ineres of shareholders and mangers and also compensaes for he leverage. In he firms wih high growh opporuniies, insiuional invesors encourage higher leverage. (3) Keiresu classificaion affecs relaions beween ownership srucure and leverage. Overall, he resuls generally suppor he hypohesis ha conflics of ineres beween shareholders and managers vary wih growh opporuniies and free cash flow and here is disciplinary role of ownership srucure in capial srucure policy. EFM classificaion codes: 110,

3 1. Inroducion The agency coss of free cash flow arise from a conflic beween managers and shareholders. When managers insulae hemselves from inernal and exernal governance mechanism, hey have incenives o pursue heir own ineress a he expenses of shareholders, e.g. higher-han-marke salaries, excessive perquisies, job securiy. Managers also end o value invesmens even if he invesmens can no maximize shareholders value since managers gain presige being he managers of a big firm. This behavior is known as overinvesmen problem in he lieraure. Jensen (1986) discussed he agency cos of free cash flow. Free cash flow is cash flow in excess of ha required o fun all projecs ha have posiive NPV. He says managers may use free cash flow o inves in negaive NPV projecs raher han reurn he free cash flow o he shareholders, for example as dividends. This problem is especially bad in firms who are maure and have few growh opporuniies, as hey have few profiable invesmens. However, by increasing deb wih is required ineres paymens, managers are bonding heir promise o pay ou fuure cash flows. Jensen indicaes ha firms wih excess cash flows and low growh opporuniies will use more deb financing for monioring purposes. Jensen and Meckling (1976) argue ha a manager, who owns a fracion, α, of he firm s ousanding equiy, bears α fracion of he cos of his divergen behavior. Jensen (1986) argues ha wase of corporae resources is an imporan insance of such behavior. As α increases, so does a manager s incenive o avoid corporae wase. This argumen implies ha he agency coss of free cash flow are decreasing funcion of α. Managers equiy ownership serves as anoher mechanism, in addiion o deb, for reducing he agency coss of free cash flow. Shleifer and Vishny (1986) argue ha large shareholders have an incenive o monior managers. Pound (1988) suggess ha insiuional invesors serve as an alernaive mechanism o conrol he overinvesmen problem. Agrawal and Mandelker (1990) indicae ha insiuional invesors provide valuable monioring services and ac as a resrain o opporunisic behavior by managers. Thus, insiuional invesors may help in reducing he firm s agency cos and become a subsiue for deb if insiuions can monior managerial aciviies a a low cos. In his sudy, we invesigae he implicaions of free cash flow heory in capial srucure policy of lised Japanese firms. The paper focuses specially on relaions beween leverage and free cash flow and growh opporuniies and how he sensiiviy of relaions varies wih ownership srucure. The ineres in sudying Japanese firms sems from hree facors. Firs, he ownership srucure of Japanese firms is quie differen from he U.S. firms. Financial -3-

4 insiuions predominae among he op five shareholders, and ake significanly larger equiy posiion in firms han hey do in he U.S. Unlike U.S. banks, Japanese banks are allowed o hold voing sock in he firms o which hey lend. Mos large indusrial firms developed a main bank relaionship wih one of he major banks. The banks mainain very close conac wih borrowers, ofen have board represenaion and are frequenly involved in major decisions or subsanial changes in company policy. In he even of financial disress, banks have virually unlimied access o corporae records and may place officers direcly in op managemen posiions. Aok Parick, and Sheard (1994) Hosh Kashyap, and Loveman(1994) and Teranishi(1993) porray he close ies beween finance and indusry in Japan as an imporan facor in he inernaional compeiiveness of modern Japanese firms. They argue ha close bank-firm relaionship can solve agency problems of managerial behavior as well as problems of asymmeric informaion beween lenders and borrowers. Some sudies sugges he exisence of negaive componens wihin he main bank sysem. Prowse (1990) suggess ha because large shareholders are also larger debholders, hey may preclude policies ha aemp o ransfer wealh from debholders o shareholders. Weinsein and Yafeh (1998) examine he effecs of bank-firm relaionship on firm performance in Japan. When access o capial markes is limied, close bank-firm ies increase he availabiliy of capial o borrowing firms, bu do no lead o higher profiabiliy or growh. The cos of capial of firms wih close bank ies is higher han ha of heir peers. This indicaes ha banks could use heir monopoly power boh o exrac significan rens from heir clien firms hrough ineres paymens, and o inhibi heir growh hrough conservaive invesmen policies. The liberalizaion of financial markes also reduces he bank s marke power. During he 1980s he Japanese financial sysem underwen a process of dramaic deregulaion and liberalizaion which made corporae borrowing and raising equiy in boh domesic and foreign capial markes much easier and more common. Wih access o foreign markes, and more developed domesic capial markes, Japanese firms gained alernaive o bank deb. This fac dramaically reduces he main bank s influence over he firm s policies and consequenly weakens he keiresu relaionship among member firms. Wih relaively weak influence over he firms policies by he main bank, managers of hese firms are relaively free o pursue self-ineres. The agen-principal conflics are likely o be a major concern o dae. Second, here exis differences beween keiresu or indusrial groups cenered around affiliaed banks and financial insiuions and unaffiliaed independen firms wih weaker banking ies. Japanese indusrial organizaion is characerized by groups of enerprises (keiresu) composed of firms based in differen indusries bu bound by ies of fracional ownership and relian on a large commercial bank as he major bu no sole lender. The large shareholders of keiresu firms ofen -4-

5 are also large crediors of he firm as well as imporan long-erm commercial business parners. The keiresu and non-keiresu firms are facing differen liquidiy consrains in heir invesmen spending. Invesmen spending is very sensiive o liquidiy consrains for non-keiresu firms, bu no so for keiresu firms. Since keiresu firms are likely o have beer access o financing sources, keiresu firms seemingly face less liquidiy consrains in making invesmen decision. The differences in insiuional arrangemens beween keiresu and non-keiresu firms may influence he behavior of shareholders as moniors. Keser (1990) describes he corporae governance sysem of keiresu firms in erms of a complex ineracion beween shareholdings, credi holding and long-erm business relaionship ha exis beween he firm and is sake holders. Aok Paric, and Sheard (1994), and Berglof and Peroi (1994) sugges a wo-ier monior sysem. In he firs sage, corporae cross-shareholders serve as he moniors under normal circumsances. Member firms are in a unique and advanageous posiion o serve as muual moniors because hey can observe each ohers performance hrough heir business relaions and because hey have specific indusry knowledge. This consan ineracion allows for beer performance evaluaion han wha can be achieved hrough cenralized monioring. Therefore, corporae muual monioring represens a crucial mechanism o ensure firms operae in a manner consisen wih all shareholders ineres. An example of he firs-sage monioring is he Presidens Council. In he second sage, he financial insiuions ake an acive inervenion role when member firms ge ino financial disress. The main bank is frequenly updaed on he financial healh of member firms hrough heir deb repaymens, while an ouside invesor has only limied abiliy o updae informaion. In addiion, Aoki and Sheard (1994) asser ha main banks serve as useful safey nes in imes of crisis because hey have he funds and experise o assume he responsibiliy. In simple cases, exising deb is refinanced and some bridge financing is provided. In more serious cases, he main bank replaces incumben managers and requires resrucuring and liquidaion of asses. Finally, he lieraure provides mixed guidance on he role of managerial ownership in he corporae governance of Japanese firms. Because he well-known keiresu srucure and influenial bank shareholders, he agency problems beween Japanese managers and shareholders are considered o be minimal (e.g. Nakaani (1984), Hosh Kashyap and Scharfsein (1990,1991) and Prowse (1990)). The manager ownership, as a way of aligning ineress beween managers and shareholders, has been viewed as an unnecessary corporae governance mechanism. This may mean greaer reliance in Japan on direc shareholder monioring, which may be refleced in he higher levels of ownership concenraion observed in Japan. (e.g. Kang and shivdasani (1995), Kaplan (1994), and Prowse (1992)). However Kang and Sulz (1998), Mock and Nakamura (1999), and Weinsein and Yafeh (1998) -5-

6 quesioned he effeciveness of bank oversigh in Japan. Morck and Nakamura (1999) argue ha for independen firms, bank equiy holders pursue heir ineress as crediors a he expense of heir equiy claims. In addiion, as banks are primarily concerned wih he firm s abiliy o mee heir deb responsibiliies, banks are less involved wih riskier firms ( Nakaani (1984) and Weinsein and Yafeh 91998)). Gibson(1995) and Kang and Sulz (1998) also argue ha poor bank healh may adversely affec heir dependen firms invesmen prospecs, which, in urn, would affec heir abiliy o monior effecively. This paricular conenion is especially relevan o he lae 1980 s and early 1990 s as i is well known ha Japanese banks have been experiencing significan financial difficulies during his ime period. In ligh of hese findings, Morck and Nakamura (1999) conend ha some independen firms may require corporae conrol mechanisms oher han bank oversigh. Due o he decline in power of Japanese banks, he rariy of incenive-based compensaion conracs for Japanese managers, and he fac ha many Japanese firms are no affiliaed wih a keiresu group, he managerial-ownership may represens an alernaive mechanism o ensure ha firms operae efficienly. Thus, he unique Japanese insiuional arrangemens provide an ineresing backdrop o invesigae wheher cash flow heory explanaion for deb policy sill apply given he differences. Tess using a sample of 833 observaions from year 1992 o 2000 suppor he free cash flow heory s predicions in he relaion beween leverage and free cash flow, growh opporuniies. The resuls also reveal ha he insiuional invesors impose heir monioring preferences hrough he governance process. Two feaures disinguish his sudy. Firs, i provides evidence consisen wih free cash flow heory predicions in a legal and regulaory environmen ha is differen form ha in he US. Second, previous researches focus only on he degree of insiuional ownership. However, he characerisics and inensiy of monioring may vary across insiuional invesors o affec corporae financing oucomes. Given he prevalence of insiuional ownership in Japanese firms, we focus on he ideniy of large block shareholders, i.e. he bank ownership, non-bank financial ownership and business corporae ownership. This sudy enhances our undersanding of he behavior of he modern corporaion. The res of he paper is organized as follows. Secion 2 presens heory and hypohesis relaing leverage, free cash flow, growh opporuniies and ownership srucure. Secion 3 explains he daa and mehodology and describes he sample characerisics. The empirical resuls are presened in Secion 4. Secion 5 makes conclusions. 2. Lieraure review and Hypoheses 2.1 Lieraure review Jensen (1986) idenified he conflic beween he shareholders ineress and he managers -6-

7 individual agendas and suggesed he deb is a remedy agains his form of agency coss: as deb forces he company o pay ou he excessive cash flow, i decreases he free cash flow which is a managers discreion and hus in danger of being subopimally invesed. Sulz (1990) shows ha i is opimal for shareholders o increase leverage when managers have personal objecives. There are a number of sudies ha invesigae he free cash flow hypohesis. One srain of empirical work examine he overinvesmen problem by analyzing he relaion beween growh opporuniies and free cash flow on he one hand, and leverage on he oher hand. They find he prediced negaive relaion beween deb and growh opporuniies (Smih and Was (1992); Lang, Ofek and Sulz(1996)) and changes in free cash flow lead o posiive changes in leverage (Cruchley and Jensen (1996)). Anoher approach o he implicaions of he free cash flow hypohesis in corporae capial srucure policy is o sudy specific evens ha afer capial srucure and show ha in general he firms aced according o free cash flow heory (Denis and Denis (1993); Blanchard, Silanes, and Shleifer( 1994)). Jensen and Meckling (1976) argue ha managerial equiy ownership reduces managerial incenives o engage in nonopimal behavior. As managerial ownership increases, managers bear more of he wealh effecs of heir divergen behaviors. Thus he disciplinary pressures of leverage and managerial ownership are subsiues. Some previous sudies find a significanly negaive relaion beween he leverage and managerial ownership(friend and Lang (1988); Jensen, Solberg, and Zorn (1992); Chen and Seiner (1999)). Shleifer and Vishny (1986) and Pound (1988) sugges ha insiuional invesors serve as an alernaive mechanism o conrol he overinvesmen problem. Insiuional invesors have greaer experise in gahering and inerpreing informaion on firms, and have more incenives o closely oversee managerial aciviies wih an increase in heir equiy ownership. This implies ha insiuional ownership and leverage serve as subsiues in conrolling managerial self-ineres if insiuional invesors impose heir managerial preferences hrough he governance process. Some evidence suggess ha here exiss a negaive relaion beween insiuional ownership and leverage. (Grier and Zychowicz (1994); Bahala, Moon, and Rao (1994); Cruchley and Jensen (1996)). 2.2 Summary of Hypoheses Based on he heoreical work of Jensen (1986) and Sulz (1990), he following hypoheses apply. Hypohesis 1: Leverage is negaively relaed o growh opporuniies and posiively relaed o free cash flow. The relaion beween free cash flow and leverage is sronger for firms wih low growh han firms wih high growh. Hypohesis 2: The leverage should be negaively relaed o managerial ownership and -7-

8 insiuional ownership in he firms wih severe overinvesmen problem. 3. Daa and Mehodology 3.1 Daa Our Japanese daa are rerieved from he Pacific-Basin Capial Markes (PACAP) daabases. The annual daa on balance shees and income saemens for lised firms we use cover 9 fiscal years from 1992 o The uiliy and financial companies are no included in he sample o avoid he effecs of governmen regulaion (Indusry Codes of , 0705 and 801 are excluded). To be included in he sample for furher analysis, we require ha he daa on equiy ownership are available. The daa on equiy ownership are aken from Japan Company Handbook. In addiion, he keiresu classificaion is obained from daabase provided by Wu and Xu (2002). 3.2 Mehodology The OLS model The hypohesis is esed using OLS regression wih Leverage raio as he dependen variable and he experimenal variables are ownership variables, Free Cash Flow, and Growh Opporuniies. The model is presened below. LEV β GROWTH 6 β PSTD 13 = β + β MANAGERIAL β ERN + β KEIRETSU β BANK + β COLVAL 8 + β NONBANK + β SIZE + β DMANUFACTURING β R & D 10 + β BUSINESS + β DMRD β SEL 11 + ε + β FCF + β NDTS Subscrip i and refer o firm and fiscal year, and year dummy variables are suppressed. ε denoes he error erms The wo-equaion model One concern over esimaing he impac of ownership srucure on leverage is he possibiliy of reverse causaion from curren leverage o curren equiy ownership. Demsez and Lehn (1985) argue ha ownership concenraions are a resul of a profi maximizing funcion, hus suggesing ha ownership srucure is endogenous o firm specific characerisics. Similarly, Cho (1998) sugges ha managers may be holding more of heir firm s shares when hey expec heir firms o perform well. This also implies ha he direcion of causaliy is such ha corporae value is affecing ownership srucure. Himmelberg, Hubbard, and Palia (1998) also provide addiional evidence ha ownership srucure is endogenous of firm-specific characerisics. In his sudy, managers are assumed o have conrol over he levels of deb and managerial ownership of equiy in he firm. Hence, he leverage raio and managerial ownership are he dependen variables. Insiuional holdings are specified as an exogenous variable since hey are exernal o -8-

9 he firm no managemen decision variable. I is deemed ha a simulaneous equaions approach is he appropriae mehodology o use. A wo-equaion model wih managerial ownership and leverage raio as he dependen variables is proposed. LEV β GROWTH 6 β PSTD 13 = β + β MANAGERIAL β ERN + β KEIRETSU β BANK + β COLVAL 8 + β NONBANK + β SIZE + β DMANUFACTURING β R & D 10 + β BUSINESS + β DMRD β SEL 11 + ε + β FCF + β NDTS MANAGERIAL = 6 BANK RSTD + 2 ADVER + 3R & NONBANK + 8 BUSINESS + 9 LEV D + + ε 4 GROWTH + 5 SIZE + Jensen, Solberg, and Zorn (1992) specify 3-equaion sysem for leverage, ownership srucure and dividends. Here we do no include a dividend equaion in our invesigaion for wo reasons. Firs, Jensen, Solberg and Zorn only finds limied evidence ha deb, insider ownership and dividends are inerdependen. Second, he agency lieraure recognizes he muliple devices o conrol agency coss such as deb policy, dividend policy, invesmen policy, incenive conracs, marke akeover, capial and produc markes. However, from an empirical sandpoin, including more endogenous variables would make he model unwieldy. In he ineres of keeping he model relaively simple, he model is confined o he capial and ownership srucure of he firm The one year lag model An alernaive mehodology o conrol for reverse causaliy is o include a one year lag beween leverage and ownership srucure and all oher explanaory variables. I also help o address he problem of spurious correlaion beween he firs version of deb raio and growh opporuniies since boh are dependen on sock price measured on he same dae. The specificaion of lagged model is given below: LEV β GROWTH 6 β PSTD 13 = β + β MANAGERIAL β ERN 1 + β KEIRETSU β COLVAL β BANK β NONBANK + β SIZE 1 + β DMANUFACTURING β R & D 10 + β BUSINESS 1 + β DMRD β SEL ε β FCF β NDTS Variable definiions LEV = oal liabiliies accouns and noe oal liabiliies + marke value of receivable equiy Managerial = percenage of shares owned by he chairman and presiden of he firm; BANK= percenage of shares owned by banks -9-

10 NONBANK= percenage of shares owned by insurance companies and russ; BUSINESS= percenage of shares owned by corporae block holders; FCFA = ( Operaing income before depreciaion-tax-gross ineres expenses on shor-erm and long erm deb-toal dividend on preferred shares-oal dividend on ordinary shares)/ Book value of oal asses; ERN = operaing income divided by sales; COLVAL = invenory, ne plan, equipmen, and land divided by book value of oal asses; SIZE = naural log of oal asses; R&D = research and developmen expense divided by book value of asses; ADVER = adverising and markeing expenses divided by book value of asses; GROWTH = (Book value of asses Book value of equiy +Marke value of equiy)/book value of asses; NDTS= depreciaion charges divided by book value of oal asses; PSTD = sandard deviaion of operaing income o sales over 5 years; KEIRETSU =1 if being keiresu member, oherwise=0; DMANUFACTURING=1 if manufacuring; oherwise=0; DMRD= 1 if R&D daa is available, oherwise=0 and reaing missing value for R&D as zero expendiure. RSTD=sandard deviaion of monhly sock reurns over a five year period. A more deailed discussion of he variables, heir measuremen, and hypohesized signs for he explanaory variables follows. Leverage Equaion Leverage. Following Prowse(1990), leverage is measured as: LEV = oal liabiliies accouns and noe oal liabiliies + marke value of receivable equiy The numeraor represens he book value of oal deb ousanding. A large porion of Japanese firms deb is ypically classified as shor-erm bank deb because he ineres rae floas wih shor-erm raes. Much of his lending carries wih i a long-erm commimen and hus is included in he measure of deb. Also, Japanese corporae balance shees presen unusually high accouns receivable and payable. To adjus for his, he excess of receivable over payable is deduced form oal deb. Following Rajan and Zingales(1995),we also ried he oher wo version of deb raio as -10-

11 following: Leverage = he sum of liabiliies/ he value of asse Leverage= (shor erm deb + long erm deb )/(asses-accouns payable-oher curren liabiliies) Ownership srucure. To measure ownership concenraion, we calculae he proporion of managerial, bank, non-bank financial insiuion, business corporaion as a percenage of ousanding shares for each firm. The percenage of shares owned by presiden and chairman is used as a proxy for managemen ownership of shares. Bank ownership is he percenage of shares owned by banks. Commercial banks for Japan refer o ciy banks, long-erm credi banks and regional banks. Non-bank ownership is he percenage of shares owned by oher financial insiuions, including pension funds, muual funds, securiies companies, and rus holdings. Business corporae ownership is he percenage of shares owned by corporae block holders; Free Cash Flow. The lieraure provides mixed guidance on he measures of free cash flow, which Jensen (1986) defines as cash flow lef afer firms have invesed all available posiive NPV projecs. Since he value of posiive NPV projecs is unobservable, free cash flow is difficul o measure in pracice. The mos commonly used FCF definiion is he one suggesed by Lehn and Poulsen(1989). Their measure of FCF is he operaing income before depreciaion minus axes, ineres expenses, and preferred and common dividends. Thus, he measure used in his sudy as a proxy for FCF is defined as: FCF = ( Operaing income before depreciaion-tax-gross ineres expenses on shor-erm and long erm deb-toal dividend on preferred shares-oal dividend on ordinary shares)/book value of oal asses ; Growh Opporuniies. The firms' growh opporuniies ha are inherenly unobservable(yoon and Sarks, 1995). The marke-o-book raio is a growh measure used mos frequenly by researchers (Smih and Was, 1992; Berger, Ofek, and Yermack, 1997; Harford, 1999). Thus, he measure used in his sudy as a proxy for growh opporuniies is he marke-o-book raio, which is defined as: GROWTH = (Book value of asses Book value of equiy +Marke value of equiy)/book value of asses We also employ several conrol variables because oher facors are known o explain leverage variaions( Harris and Raviv,1991). The variables we use in regression analyses and heir heoreical jusificaions are presened below. Profiabiliy. Several sudies find a negaive relaion beween leverage level and profiabiliy( Jensen, Solberg, and Zorn(1992), Keser (1986), Timan and Wessels (1988). The reason is ha more-profiable firms are in a posiion o rely on inernally generaed funds(myers -11-

12 and Majluf, 1984). We use he raio of operaing income o sales as a proxy for profiabiliy. Collaeral values. Secured deb gives lenders ile o pledged asses unil deb is fully paid. Because crediors can ake possession of fixed asses, hey can avoid he legal processes ha, in he even of bankrupcy, are required for unsecured crediors. This implies a posiive relaion beween deb measures and he capaciy of firms o collaeralize heir deb. Keser (1986), Prowse(1990), Rajan and Zingales (1995) and Wald (1999)provides empirical suppor for his relaion in Japan. We use he sum of invenory and ne plan, equipmen, and land over oal asses as measure of collaeral asses in he firm. Firm Size. Larger firms are argued o be more diversified and less prone o bankrup han smaller firms. As a resul, hey have he abiliy o suppor more deb. Jensen, Solberg, and Zorn (1992) find empirical suppor for his hypohesis. In addiion, Hodder and Tschoegl(1985) argue ha he Japanese governmen disproporionaely direcs funds o large firms, who borrow heavily. These conenions sugges a posiive relaion beween firm size and leverage. We use he naural logarihm of oal asses as measure of size. R&D. Myers (1977) presen a model in which ousanding deb causes underinvesmen in fuure opporuniies. If he firm has capial enrenched in physical plan, he poenial for underinvesmen or excessive risk aking by managemen is reduced. Thus, measures of physical plan are expeced o increase he deb raio of a firm. Symmerically, firm s expense on inellecual goods, such as R&D, are expeced o decrease he deb raio. R&D may carry addiional ax deducions. The coefficien should be negaive o deb raios. Wald (1999) find negaive coefficien of R&D. ADVER. Timan (1984) argues ha he firm s liquidaion imposes coss on cusomers and employees. Since firms wih unique producs impose relaively high coss from liquidaion on cusomers and employees, uniqueness of he asss is expeced o be negaively relaed o leverage. Uniqueness of he asses is measured by R&D and ADVER. Thus a negaive sign is expeced. Non-deb ax shields. DeAngelo and Masulis(1980) sugges ha various forms of nondeb ax shields, including depreciaion charges, can subsiue for corporae deb. hus, oher hings being equal, hey expec a negaive relaion beween deb levels and he availabiliy of nondeb ax shields. They sugges ha firms ha inves heavily in angible asses have high depreciaion deducions, which, in urn, allow hem o use less deb. Keser (1986) provides empirical suppor of his negaive relaion for Japan. We use he raio of depreciaion charges o oal asses. Volailiy of Profiabiliy. High operaing risk increases he probabiliy and cos of bankrupcy. Therefore, high-operaing-risk firms end o use less deb han low-operaing-risk firms(bradley, Jarrell, and Kim,1984; Jensen, Solberg, and Zorn,1992). We employ he sandard deviaion of he previous five years operaing income over sales. -12-

13 Keiresu. This is a dummy variable equal o zero for independen firms and one for keiresu firms. This index is deermined from he group s influence power, which is he raio of group s shareholding o he oal shares held by he op en shareholders. Keiresu membership is readily idenifiable and says raher sable over ime he swich-in and -ou of membership (especially when a weak membership is no couned) is very rare. Following Wu and Xu (2002), keiresu firms are he group firms from he classificaion by Nakaani (1984) or he closely affiliaed members (wih he degree of he 2-, 3-, and 4-sar inclinaions) of he six keiresu groups from he classificaion in he 1992/93 ediion of Indusrial Groupings in Japan he Anaomy of he Keiresu. Oher firms ha do no fall in eiher of he above classificaions belong o non-keiresu firms. Nakaani(1984) sugges ha ha keiresu firms are associaed wih higher levels of deb because of he smaller risk of bankrupcy for keiresu firms since banks are he principle shareholders of hese firms. Indusry classificaion. According o Timan(1984), a firm s liquidaion can impose coss on is cusomers if i makes producs ha require mainenance in he fuure. However, cusomers ake ino accoun he liquidaion likelihood and reduce demand and prices of he producs, as a resul, companies in cerain indusries, such as hose ha produce machines and equipmen, may have high indirec bankrupcy coss and, herefore, have o bond hemselves by choosing capial srucures wih relaively low leverage. Timan and Wessels (1988) provide empirical suppor for his conenion. To examine unique indusry effecs, we include a series of dummy variables o idenify indusry classificaion. Firms in manufacuring indusries are equal o 1, oherwise equal o 0. Managerial Ownership Equaion RSTD is he sock reurn volailiy measured as he sandard deviaion of monhly sock reurns of he firm over a five-year period. Because heir human capial is ied o he firm, managers would be relucan o commi heir personal wealh o he firm as his grealy diminishes heir diversificaion opporuniies. Furher, his non-diversificaion problem becomes more severe as he sock reurn volailiy of he firm increase. To capure his effec, he sandard deviaion of monhly sock reurns esimaed over five years is used o proxy for oal risk. Based on he argumen pu forh, a negaive coefficien for PSTD is hypohesized. R&D and Adverising expenses variable is defined as before and capures he discreionary invesmens of he firm. Since hese invesmens are characerized by greaer agency coss (Myers, 1977), R&D and Adverising inensive firms should be associaed wih greaer insider ownership in equilibrium. Consequenly, a posiive coefficien is hypohesized for he R&D and Adverising expenses. Growh represens oal asse growh and is anicipaed o have a posiive influence on -13-

14 managerial ownership. To he exen ha pas asse growh reflecs fuure profiabiliy and growh poenial, managers would be less relucan o inves in he firm s equiy. The posiive relaionship beween growh and managerial ownership could sem from informaional advanages o insiders abou growh prospecs of he firm. Managers would be more inclined han he average invesor o be on he growh prospecs because of heir beer knowledge of projecs being underaken by he firms. Size is sill measured as he log of oal asses of he firm. previous sudies have documened ha managerial ownership is much greaer in smaller firms han in larger firms. As he firm ges larger, a relaively smaller proporion of shares are owned by managers owing o heir limied personal wealh, consrains on personal borrowing, and he diversificaion problem. Therefore, a negaive coefficien for size is expeced. Bank, non-bank and business corporae ownership are all insiuional ownership. The subsiuabiliy argumens sugges ha he opimal proporion of managerial ownership should decline wih increasing insiuional monioring. A negaive coefficien is expeced for insiuional ownership. Leverage is he same as before. The subsiuabiliy argumen suggess an inverse relaionship beween managerial ownership and he deb raio. A negaive coefficien for Leverage is anicipaed. Year dummies for each year are included in regressions o capure unobserved macroeconomic effecs The addiional es for keiresu members Previous empirical sudies sugges ha here exi wo disinc sysems of corporae governance beween he independen firms and he keiresu members. The independen firms are much more similar o U.S. firms in heir arms-lengh relaionships wih financiers and rading parners. The weaker ies ha exis beween independen firms and heir sakeholders sugges ha mehods of monioring and influencing managemen may rely more direcly on he size of he shareholder s equiy holdings, as is he case in he U.S. Hence furher es are conduced o examine he role of keiresu vs. non-keiresu firms. 3.4 Sample Selecion For each year, firms are ranked according o heir fiscal year-end M/B. The boom quarile is seleced o represen firms wih low growh opporuniies. The op quarile is seleced o represen firms wih high growh opporuniies. To be included in he sample for furher analysis, we require he observaions saisfied several condiions: 1) he managerial ownership is available. If he manager ownership is no on he op

15 shareholders lis in he Company Handbook, he observaion will be excluded. To ake ino accoun he possible confounding effecs of managemen (direcor) enrenchmen, he observaions wih managerial ownership greaer han 50 percen will also be excluded. According o Shleifer and Vishny(1989), he correlaion beween managemen ownership and agency coss is U shaped. In oher words, agency coss decline from low o moderaed levels of managemen ownership as a resul of he alignmen of ineres beween managers and shareholders. However, a very high levels of managemen share ownership, agency coss increase as a resul of managemen enrenchmen. An analysis of his relaion is beyond he scope of his sudy. Therefore, all observaions wih direcor ownership greaer han 50 percen are excluded. 2) The observaions have no missing value for Leverage Raio, Free Cash Flow, and sandard deviaion of operaing income o sales over 5 years. 3) For he 2-equaion sysem model, he observaions should have no missing value of he sandard deviaion of monhly sock reurns over a five year period. A breakdown of low-growh and high-growh firms by year, indusry, and Keiresu classificaion is provided in Table-1. Panel A shows ha he low-growh sample conains 398 observaions for 137 firms, while he high-growh sample consiss of 435 observaions for 153 firms, respecively. Panel B of able 1 indicaes indusry disribuion of he sample. The low-growh sample conains 6 indusry groups, and firms in manufacuring indusries (indusry codes beween 301 and 315) represen percen of all firms. The high-growh sample conains 5 indusry groups, and firms in manufacuring indusries represen percen. Panel C of Table 1 shows he Keiresu and Independen Firms in he sample. The low-growh sample conains 40 keiresu members and 395 independen firms. The high-growh sample conains 30 keiresu members and 368 independen firms. 3.5 Descripive saisics Table-2 provides descripive saisics for low-growh and high-growh firms. The firs column idenifies he variables, he second and hird columns provide descripive saisics for each subsample, and he las column repors he p-value for he difference in means (median) beween he wo subsamples. As expeced, high-growh firms have lower amoun of deb in heir capial srucure han low-growh firms. The mean (median) leverage for high-growh firms is 20.2 percen (18.7 percen), compared o a mean (median) of 68.2 percen (71.5 percen) for low-growh firms. Boh he -es and he Wilcoxon es indicae ha he difference beween he wo subsamples is significan (p<0.01). This is consisen wih he free cash flow hypohesis in which firms wih -15-

16 fewer growh opporuniies have higher levels of leverage. The difference in he Growh Opporuniies (M/B raio) beween he wo subsamples is significan (p<0.01). The mean (median) M/B for he high-growh subsample is (2.065), compared o a mean (median) of (1.144) for he low-growh subsample. There is also evidence ha ownership srucure differs beween he wo subsamples. Low-growh firms have lower managerial ownership and non-bank, business corporae ownership bu higher bank ownership han high-growh firms. The mean (median) managerial ownership for low-growh firms is 6.7 percen (4.4 percen), compared o a mean (median) of 8.6 percen(6 percen) for high-growh firms. The mean (median) non-bank ownership for low-growh firms is 6.5 percen (5.8 percen), compared o a mean (median) of 7.9 percen (6.4 percen) for high-growh firms. The mean (median) business corporae ownership for low-growh firms is 10.1 percen (5.2 percen), compared o a mean (median) of 11.9 percen (8.1) percen for high-growh firms. On he oher hand, he mean (median) bank ownership for low-growh firms is 9.1 percen (9 percen), compared o a mean (median) of 8.3 percen (8.6 percen) for high-growh firms. Based on difference in means, managerial and business corporae ownership differ significanly (p<0.01) beween he wo subsamples. Bank and non-bank ownership differ less significanly (p<0.05). 4. Regressing Resuls The regression analyses are conduced in hree sages. The firs sage uses pooled cross-secional ime-series daa and regresses he same year dependen variable by OLS. The second sage regresses he same year dependen variable by 2SLS. The hird sage regresses he lagged dependen variable by OLS. Finally, furher ess are conduced o examine he role of Keiresu vs. non-keiresu firms in corporae finance policy. 4.1 The resuls of OLS model. Table 3 presens he resuls for relaions beween leverage and ownership srucure, Free Cash Flow and growh opporuniies. The second and hird column provide regression resuls for low-growh and high-growh firms. The values of -saisics are adjused for he heeroskedasiciy in residuals using he Whie(1980) mehod, and are presened in parenhesis below each coefficien esimae. The resuls show a significan negaive relaion beween leverage and growh opporuniies. The coefficien for growh opporuniies for low-growh and high-growh firms are and , respecively, and saisically significan. The coefficien for low-growh firms is much lower han ha for he high-growh firms. This findings lend suppor o he Jesen s (1986) argumen ha firms wih few profiable invesmens are bonding heir promise o pay ou fuure cash -16-

17 flows by increasing deb. Resuls sugges ha firms wih greaer agency coss of free cash flow use more leverage. The coefficiens on free cash flow for low-growh and high-growh firms are and 0.756, respecively, and saisically significan. This is consisen wih Jensen (1986). There is also evidence ha he sensiiviy of relaions beween ownership srucure and leverage varies wih growh opporuniies and Free Cash Flow. The coefficiens on managerial ownership for low-growh and high-growh firms are and 0.010, respecively, and no significan. The coefficien on he bank ownership for low-growh and high-growh firms are and 0.395, respecively, and significan (p<0.01). The coefficien on non-bank ownership for low-growh and high-growh firms are and 0.180, respecively, and significan(p<0.10). The coefficien on business corporae ownership for low-growh firms is and no significan, while he coefficien for high-growh firms is and significan(p<0.01). These findings provide suppor o Hypohesis 2 ha divergence of ineres beween shareholders and managers over capial srucure is greaer for low-growh firms, and herefore he role of equiy ownership in miigaing he overinvesmen problem is more imporan for hese firms. In his kind of firms, leverage declines as managerial ownership, bank ownership, non-bank ownership and business corporae ownership rise. As suggesed by Jensen and Meckling (1976), managerial ownership aligns he ineres beween managers and shareholders. Firms wih high managerial ownership have lower agency coss of free cash flow, hey need less leverage o conrol he agency problem. However, when he bank and non-bank financial insiuions provide valuable monioring services and ac as a resrain o opporunisic behavior by managers, firms find i opimal o use lower level of leverage, which is an alernaive mechanism o reduce he agency conflics. For high-growh firms, which has less agency cos of free cash flow, here is no compensaion relaion beween leverage and insiuional ownerships. The insiuions are responsible for he higher leverage raio. While mos of he conrol variables have signs as expeced based on heories of capial srucure. The coefficien for profiabiliy is negaive. I suppor for he pecking order heory by Myers and Majluf(1984). Consisen wih Bradley, Jarrell, and Kim(1984), we find a posiive relaion beween deb and collaeral values and negaive relaion beween deb and earning volailiy. For he low-growh firms, he posiive sign for size is consisen wih Jensen, Solberg, Zorn(1992) and Hodder and Tschoegl(1985) and shows ha he abiliy o suppor more deb. The negaive coefficien for R&D suppors for Myers (1977) hypohesis ha R&D-inensive firms, incur high agency coss because of heir greaer discreionary invesmen. The negaive coefficien for R&D is also consisen wih BJK s finding. For he low-growh firms, he negaive coefficien suppors Timan s (1984) argumen ha he uniqueness of he asss is expeced o be negaively relaed o -17-

18 leverage. The depreciaion reflecing non-deb ax shield is wih a negaive sign, which provide suppor for DeAngelo and Masulis (1980). The coefficien for Keiresu dummy is negaive, which show keiresu firms have higher deb raio han non-keiresu firms. The resul is consisen wih he finding of Nakaani(1984). Finally, indusry dummy variable exhibi significanly posiive signs. This is consisen wih Timan s (1984) conenion ha firms in nonunique indusries end o have higher leverage han hose in unique indusries. Two excepions is he Size and Adverising expenses for high-growh firms. The coefficien on Size for high-growh firms are bu insignifican. The coefficien on adverising expenses for high-growh firms is and significan. A possible explanaion is ha he adverising adds value o he firm, which increases borrowing capaciy. 4.2 The resuls of wo-equaion model Table 4 presens he esimaes of he simulaneous wo-equaion sysem. For he deb equaion, he signs and significance of he coefficiens using sysem equaion are similar o he OLS esimaes. For he low-growh firms, he coefficien on bank ownership, free cash flow and growh opporuniies are , 1.20 and respecively, and significan. For he high-growh firms, he coefficien on business ownership, free cash flow and growh opporuniies are 0.178, and respecively. The coefficien on free cash flow is insignifican. The resuls show he low-growh firms, wih greaer agency coss of free cash flow, use more leverage and he bank provide valuable monioring services and ac as a resrain o opporunisic behavior by managers. The agency cos of free cash flow is no severe in high-growh firms and business ownership is responsible for is high leverage raio. Wih respec o he Managerial ownership equaion, mos of he signs of he coefficiens are in he direcion hypohesized. As prediced, he bank ownership, non-bank ownership and business ownership has a significan negaive coefficien as hypohesized. I was argued ha insiuional ownership fosers addiional monioring and acs as a resrain o he opporunisic behavior on he managers. The negaive coefficien for R&D is consis wih he Myers hypohesis ha R&D-inensive firms are associaed wih greaer agency cos of deb making managerial ownership a relaively more imporan device in conrolling he agency problem of he firm. For he high-growh firms, he posiive coefficien of Growh is consisen wih managers preference o inves in he firm s equiy if fuure prospecs are good. This coefficien for low-growh firms is negaive, bu insignifican. The negaive coefficien for size is consisen wih previous sudies ha documen a lower proporion of managerial ownership in larger firms owing o limied personal resources. The excepion is he volailiy of sock reurn. The sign is significanly posiive, bu i is near The resuls of Lag model -18-

19 The resuls for regression using lagged dependen variable for he year 1992 o 1999 are repored in Table 5. As noed, hese resuls are far more meaningful han he OLS given he lag beween he idenificaion of growh opporuniies and he dependen variable. The resuls show ha for he low-growh firms, he free cash flow is posiively relaed wih he leverage. The growh opporuniies is negaively relaed wih he leverage. The bank ownership is negaively relaed wih he leverage. For he high-growh firms, he free cash flow is negaively relaed wih he leverage, he growh opporuniies is negaively relaed wih he leverage. The bank ownership and business ownership are posiively relaed wih he leverage. All he resuls are significan. 4.4 Addiional analysis for keiresu classificaion Descripive saisics Table 6 shows he comparison of firm characerisics for he four kinds of firms. The column 1 idenifies he variables used in he sudy, he column 2 hrough 5 provide saisics for he wo subsamples compared. The column 6 o 7 repors he p-value for he difference of means es beween he wo kinds of firms. The column 6 is based on he -es and he column 7 is based on he Wilcoxon wo-sample es. If boh p-values are less han 0.01, we idenify he difference as significan. The panel A compares he low-growh keiresu firms vs. low-growh independen firms. The difference in he non-bank ownership is significan. The mean (median) of non-bank ownership for low-growh keiresu firms is 0.11(0.108), compared o 0.062(0.053) independen firms. The panel B compares he high-growh keiresu firms vs. high-growh independen firms. The difference in he business corporae ownership is significan. The mean (median) of business corporae ownership for high-growh keiresu firms is 0.154(0.109), compared o 0.116(0.08) for independen firms. Panel C compares he low-growh Keiresu firms and high-growh Keiresu firms. The differences in leverage, bank ownership, business ownership, and growh opporuniies are significan. The mean (median) of leverage for low-growh Keiresu firms is 0.633(0.629), compared o 0.251(0.28) for high-growh firms. The mean (median) of bank ownership for low-growh Keiresu firms is 0.11(0.114), compared o 0.072(0.085) for high-growh firms. The mean (median) of business corporae ownership for low-growh Keiresu firms is 0.057(0.045), compared o 0.154(0.109) for high-growh firms. The mean (median) of growh opporuniies for low-growh Keiresu firms is 1.223(1.28), compared o 2.129(2.052) for high-growh firms. Panel D compare he low-growh independen firms and high-growh independen firms. The differences in leverage, non-bank ownership, free cash flow, growh opporuniies are significan. The mean (median) of leverage for low-growh independen firms is 0.686(0.725), compared o 0.197(0.179) for high-growh firms. The mean (median) of non-bank ownership for low-growh -19-

20 independen firms is 0.062( ), compared o 0.078(0.059) for high-growh firms. The mean (median) of free cash flow for low-growh independen firms is 0.022(0.017), compared o 0.044(0.044) for high-growh firms. The mean (median) of growh opporuniies for low-growh independen firms is 1.163(1.142), compared o 2.293(2.065) for high-growh firms Regressing resuls Furher ess are conduced o examine he role of Keiresu vs. Non-keiresu firms. As in he earlier case, hree se of regression are run; using he OLS, 2-SLS and lagged dependen variable. The resuls are repored in Table 7, Table 8, and Table 9. Here, we only discuss he resuls of he OLS in deail since he resuls of 2SLS and Lagged dependen variable are maerially idenical. For he independen firms, he resuls show he associaion beween leverage and free cash flow, growh opporuniies, and he impac of he ownership srucures are similar o he earlier resuls. For he low-growh independen firms, he coefficien on free cash flow is (p<0.10), much high han (p<0.05) for high-growh firms. The coefficien on growh opporuniies is (p<0.01), much lower han (p<0.01) for high-growh firms. For low-growh independen firms, he relaion beween leverage and managerial ownership and insiuional ownerships are negaive and he coefficien of bank ownership, and non-bank ownership are significan. For he high-growh independen firms, leverage is posiively relaed wih managerial ownership and insiuional ownerships and he coefficiens of bank, non-bank ownership and business ownership are significan. The resuls is consisen wih he free cash flow hypohesis in which firms wih fewer growh opporuniies have greaer agency coss of free cash flow and use more leverage. For he low-growh independen firms, he bank invesors may impose heir managerial preferences hrough he governance process, e.g. he bank invesors seek o increase heir oversigh on firms hrough sending heir represenaives o si on he boards. Hence, he bank ownership and leverage serve as subsiue in conrolling he managerial self-ineres. For he high-growh independen firms, he non-bank and business corporae invesors exercise heir influence hrough eliciing higher leverage o conrol he overinvesmen problem. One possible explanaion is he difficulies for he insiuional ownership o gaher and inerpreing informaion on his kind of firms because of he need of specific indusry knowledge since hey are no affiliaed wih he keiresu hrough business ies and reciprocal shareholdings. These difficulies deer he insiuional invesor o conrol he agency problem hrough governance process. For he keiresu firms, he resuls are mixed. For he low-growh keiresu firms, he coefficiens on he free cash flow and growh opporuniies are and respecively, and insignifican. The relaion beween he leverage and non-bank ownership and business ownership are posiive and significan, which show he bank and business corporae invesors elici higher -20-

21 leverage for he low-growh keiresu firms. The resul of low-growh firms doesn lend much suppor o he Hypoheses of free cash flow. One possible explanaion is ha his kind of firms rely monioring mehods oher han leverage and ownership srucure o conrol he agency problems. For he high-growh keiresu firms, he coefficien on free cash flow and growh opporuniies are (p<0.01) and (p<0.05) respecively, compared o he (insignifican) and (p<0.01) of high-growh independen firms respecively, which show he agency problem of free cash flow is more severe in he keiresu members due o heir easy borrowing. The relaion beween he leverage and non-bank ownership is negaive and significan, which show he insiuional ownership compensae for he leverage in monioring he problem of free cash flow. The relaion beween he leverage and managerial ownership is and significan a 10.96% ( significan a 9.95% in he sysem regression), which show he managerial ownership compensae for he leverage in his kind of firms. 5. Furher es on he Low-growh high free cash flow and High-growh low free cash flow sub-sample. According o he free cash flow heory, he divergence of ineres beween shareholders and managers should be he mos severe in he firms wih few growh opporuniies and large free cash flow. Hence he relaions beween ownership srucure and leverage are mos imporan for hese firms. To invesigae his argumen, sample of low-growh and high-growh firms are classified as he low-growh high free cash flow firms, high-growh low free cash flow firms, and inermediae firms. Firms wih free cash flow greaer han he median value are placed in he high free cash flow firms, and ones wih less han he median value placed in he low free cash flow firms. Low growh high free cash flow firms are mos likely o have he overinvesmen problems, while high growh low free cash flow firms are leas likely. Oher firms are placed in inermediae firms, wih growh dummy=1 if hey are from he high-growh group. Table 10 presens he resuls of hree kinds of firms. The coefficien of free cash flow for he low-growh high free cash flow firms is he highes in he hree subsamples. The coefficien of free cash flow for he high-growh low free cash flow is he lowes in he hree subsamples. The coefficien of growh opporuniies for he low-growh high free cash flow firms is he lowes in he hree subsamples. The coefficien of growh opporuniies for he high-growh low free cash flow is he highes in he hree subsamples. The coefficien of ownership are all negaive for he low-growh high free cash flow firms and ha of bank ownership is significan (p<0.01). The coefficiens of ownership are all posiive for he high-growh low free cash flow firms and he coefficiens of bank and nonbank ownership are significan (p<0.05). The resuls indicae ha he conrol devices of leverage and ownership srucure are mos imporan for Low-growh high free -21-