Capital Structure, Firm Liquidity and Growth

Transcription

1 Capial Srucure, Firm Liquidiy and Growh Ronald W. Anderson 1 May Inroducion This paper is an exploraion of he relaionships among he firm's financial srucure, is choice of liquid asse holdings, and growh. The lieraure on he deerminans of economic growh has long sudied he role of financial developmen in promoing growh. 2 The lieraure on corporae finance provides some microeconomic underpinnings for his relaionship in he form of a variey of parial equilibrium analyses showing how capial srucure can effec he firm s invesmen decisions and herefore is growh. 3 To dae his lieraure has overlooked he channel beween financial srucure and growh ha we examine here. Briefly, he cenral hesis of our paper is ha in choosing o hold is asses in liquid form, he firm will ofen forego an opporuniy o inves in an illiquid and, possibly riskier, projec ha promises a higher expeced reurn. In so doing, he growh of he firm is slowed since high reurns on asses will pave he way for fuure invesmens. The decision o hold asses in liquid form may be moivaed by a variey of consideraions. Prominen among hese is ha liquid asses may allow he firm o inves in a more aracive growh opporuniy ha may presen iself laer (i.e., he opion value of waiing o inves is high). The alernaive reason ha we focus on is a precauionary moive for keeping a high level of liquidiy. Liquid asses provide a cushion ha would allow he firm o survive a period of low earnings during which he firm migh be unable o access capial markes or could do so only a a very high cos. The firm s financial srucure will affec his decision because he degree of leverage used by he firm will affec he likelihood ha cash flows will be insufficien o cover deb service and oher fixed charges. This creaes a possible linkage from high deb o high liquidiy o slow growh. We explore his idea in hree way in wha follows. Firs, we presen he findings of a heoreical model of he firm where exernal finance is cosly and where reaining earnings as 1 IRES, Universié caholique de Louvain and Deparmen of Accouning and Finance, London School of Economics and Poliical Science. The paper repors research carried ou wihin he Naional Bank of Belgium's projec, "New Views on Firms' Invesmen and Finance Decisions". I would like o hank he NBB for heir suppor and my co-researchers on his projec for heir commens. I would also like o hank my colleague Andrew Carverhill of Hong Universiy of Science and Technology for discussions and for allowing me o draw upon our join heoreical research. Elizabeh Mueller has provided excellen research assisance. Responsibiliy for all errors and views expressed is my own. 2 This lieraure goes back a leas o Schumpeer (1942). More recenly he heme has been explored in heoreical models of growh, e.g., Greenwood and Jovanovich (1990) and in he empirical lieraure on growh, which has documened posiive correlaions beween growh raes and various measures of financial developmen, e.g., King and Levine (1993). 3 For a recen conribuion including a review of he earlier corporae finance lieraure see Anderson and Nyborg (2002).

2 liquid asses serves a precauionary moive. We sudy he policy ha should be chosen by managers maximizing shareholder wealh and discuss is implicaions for dividend policy and liquid asse holding. One of he predicions of his model is ha a long-erm reliance on high levels of deb finance ends o be associaed wih high levels of liquid asse holding. We hen sudy he quesion empirically by esimaing he deerminans of liquid asse holdings using panel daa ses of Belgian and UK firms. Afer conrolling growh opporuniies and oher imporan facors we find evidence of a posiive relaion beween leverage and liquid asse holding. This is consisen wih he heoreical analysis we presened and goes agains previous sudies, which documened a negaive relaionship beween leverage and levels of liquid asse holding. Finally, we discuss he implicaions of hese resuls from a broader insiuional perspecive. In paricular, we discuss he possible implicaions of he developmen of sock markes, privae equiy, and venure capial markes. The remainder of his paper is organized as follows. In secion 2 he previous lieraure is reviewed. In secion 3 we presen he resuls of a heoreical analysis of a dynamic model of firm liquid asse holding. Secion 4 is devoed o our empirical analysis. Secion 5 discusses implicaions of he insiuional developmen in ligh of our findings. Secion 6 concludes. 2. Lieraure Review The finance lieraure on he corporae choice of liquid asse holding is no very large. Unlike he enormous lieraure on capial srucure, he composiion of he lef hand side of company balance shees has no araced a grea deal of aenion. The reason for his is in par aribuable o he classic proposiions of Modigliani and Miller esablishing he irrelevance of financial srucure and dividend decisions for he value of he firm and he cos of capial. For firms operaing in an M&M world he asses of he firm will be composed of hose invesmens available o he firm for which he ne presen value is posiive. Firms can inflae heir balance shees by making zero NPV invesmens in liquid asses; however, here seems lile reason o believe ha firms would wish o do so on a sysemaic basis. Tha is, here will be no relaionship beween he firm s preference for liquidiy and capial srucure or oher firm characerisics. The enormous developmens of he lieraure on corporae finance in he las weny-five years have of course produced a plehora of ideas as o how capial srucure does maer for he value of he firm and for he firm s invesmen decisions. Poenially, agency effecs of various kinds may creae imporan reasons for holding liquid asses wih he furher implicaions of differen paerns of corporae liquidiy depending on capial srucure or oher firm characerisics. This was one of he messages of Myers s original analysis of he corporae borrowing decision (Myers (1977)). By poining ou ha problems of deb overhang (i.e., levered equiy s disincenive o inves in posiive NPV projecs) could be overcome wih financial slack he esablished an imporan incenive for mainaining liquid asse holding. This noion has been furher reinforced by he subsequen corporae finance lieraure ha has emphasized he reasons why apping capial markes may be relaively cosly compared o reaining earnings. 4 Furhermore, he lieraure on real opions has 4 This idea was developed more explicily in Myers and Majluf (1984). I is also sudied in his volume in he paper of Van Cayseele. 2

3 emphasized he opion value of waiing o inves 5. Thus mainaining liquid asses will be imporan for firms faced wih growh opporuniies whose expeced reurn flucuaes over ime. Oher veins of he corporae finance lieraure have poined oward possibly unfavourable effecs of liquid asse holding. This is an implicaion of asse subsiuion problem firs idenified by Jensen and Meckling (1976). If levered equiy holders have an incenive o engage in increasing he riskiness of asses, hey can implemen his only if he firm s asses are sufficienly liquid o allow his ransformaion o ake place. These posiive and negaive aspecs of corporae liquidiy have been sudied ogeher in a recen analysis by Myers and Rajan (1998). Asse liquidiy besows a benefi in he sense ha i makes i possible for he firm o seize unforeseen opporuniies or o survive during periods of poor business condiions. However, liquidiy has a disadvanage ha i makes i difficul for firm insiders o commi o a given course of acions. Their analysis assumes ha asses in place generae cash flows a daes =1 and =2. They allow insiders o underake non-conracible asse ransformaions a daes =0.5 and =1.5. The values of ransformed asses are supposed o be appropriaed enirely by firm insiders o he derimen of he firm's crediors. Asse liquidiy is he proporion of he asse value ha is reained upon ransformaion. They show ha he deb capaciy of he firm i ypically non-monoonomic firms wih excessively liquid can raise less deb han can a less liquid firm. Rajan s and Myers model is very sylised and heir analysis is incomplee in many respecs. However, hey clearly idenify a conflic of ineress of debholders and shareholders in regard o he degree of liquid asse holding. Recenly, Morellec (2001) provides an ineresing analysis of he implicaions of liquidiy ha develops he asse ransformaion heme. He considers a dynamic model of a levered firm whose ne revenue, given a capial sock, follows a geomeric Brownian moion. The firm benefis from leverage because of he ax shields hey besow. He sudies he firmclosure decision of a levered firm and obains closed-form expressions for he values of deb and of equiy. He hen considers he possibiliy ha he firm can liquidae a fracion of he asses of he firm prior o closing he firm enirely. He shows ha shareholders will generally do so and ha given a level of nominal deb, his parial asse sale increases he value of equiy and reduces he value of deb. Thus, ex ane, asse liquidiy (i.e., he capaciy o engage in asse sales) reduces he value of he firm and he deb capaciy of he firm. Consequenly, asse liquidiy can resul in underinvesmen relaive o he illiquid asse benchmark case. Kim, Mauer and Sherman (1998) presen a heory of he choice of liquid asses when ouside finance is cosly. The basic moivaion for holding liquid asses is he creaion of financial slack which will allow firm insiders o pursue fuures aracive invesmen opporuniies when hey presen hemselves (as in Myers and Majluf). They sudy his issue in a hree period model. A ime zero a firm wih amoun of funds available, X allocaes hese funds o an illiquid invesmen wih a risky reurn, a liquid invesmen wih a (low) riskless reurn, and dividends. A =1, given he reurns o iniial invesmens, he shareholders issue, a a cos, an amoun of deb and chooses a level of risky fixed invesmen. A ime =2 reurns are 5 For developmens of his idea and references o he lieraure see Cassimon e.al., Buzen e.al. and Gérard and Verschueren all in his volume. 3

4 paid our as a dividend and firm is dissolved. In his se-up we see ha he moivaion for holding liquid asses is o avoid issuing cosly deb a ime =1. Kim, Mauer and Sherman explore heir heory in an empirical sudy of U.S. indusrial firms over he period from 1975 hrough Their measure of liquidiy is he raio of cash and markeable securiies o he book value of asses. They consider explanaory variables measuring growh opporuniies, cash flow volailiy, deb raio, cash flow, and bankrupcy risk. They inerpre heir findings as supporive of heir heoreical model. This is based principally upon he finding of a posiive relaionship beween growh opporuniies and liquidiy. I is noable ha hey find a negaive relaionship beween leverage and corporae liquidiy for which heir heoreical analysis offers no explanaion. We will discuss heir resuls again in Secion 4 when we urn o our own empirical analysis. Opler, Pinkowiz, Sulz and Williamson (1999) also provide an empirical analysis of corporae holding of liquid asses. Their sample covers U.S. non-financial firms during he period. Their resuls are similar o hose of Kim e.al. In paricular, hey find a posiive relaionship wih respec o growh opporuniies (measures by marke o book raio or by R&D spending). One difference is ha Opler e.al find ha cash flow eners posiively; whereas, Kim e.al. find a negaive relaion wih cash flow. In agreemen wih Kim e.al., hey find a negaive relaion beween deb and corporae liquidiy. We will reurn o his and oher aspecs of heir analysis in Secion 4 below. 3. The Precauionary Moive for Corporae Liquidiy 6 In his secion we address he quesion: in a world where a firm may be exposed o persisen periods of relaively low (or high) cash flows and where accessing exernal capial markes is very cosly wha policy for paying ou dividends or reaining earnings in he form of liquid asses will be in he ineress of shareholders? Consider a firm wih given producive capaciy (asses in place). The firm may accumulae liquid asses as a reserve o service deb and oher fixed flow coss. To capure in a simple way he idea ha ouside finance is cosly we simply do no allow for new equiy issues or issue of new deb. Asses in place generae cash flows following a diffusion wih sochasic drif. Liquid asses have a reurn less han he rae used o discoun fuure profis, and less han he mean drif of he firm's risky asses. Mahemaically ds is he cash flow generaed by asses in place, r is he discoun rae, I is flow coss including deb service (i.e., deb is considered o be a perpeual bond wih fixed ineres). Then he echnology is S ds = ρ d + σdw S where ρ is a sochasic drif erm, σ is cash flow volailiy (a consan) and dw is a Brownian moion. In order o reflec persisence in deparures from normal levels of cash flows we model he drif wih mean reversion according o he Ornsein-Uhlenbeck process, d ρ = κ ( ρ ρ ) d + ηdw ρ 6 This secion is based on unpublished join work wih Andrew Carverhill of Hong Kong Universiy. 4

5 ρ where dw is independen of discouned value of dividends. S dw. Shareholders are risk neural and maximize he presen The difference beween cash flows and cash ouflows in he form of coss, deb service and dividends is accumulaed in liquid asses belonging o he firm. So long as he cash flows plus liquid asses are sufficien o cover coss and deb service he firm is solven. Oherwise he firm is bankrup. In he case of bankrupcy he firm is insananeously ransferred o bondholders and he firm subsequenly is operaed as an all equiy firm. We assume ha here are no bankrupcy coss. We furhermore assume ha he reurn on liquid asses is zero; however, his is jus a convenience. Our resuls hold rue if reurns o liquid asses held by he firm is bounded above by he risk-free rae. 7 In his se up here is no accumulaion of risky asses, so ha he firm's only invesmen choice concerns he amoun of liquid asses which are o be held. Therefore, he cos of holding liquid asses is he pure ime value of money (r), i.e., he loss associaed wih consuming laer raher han earlier. The benefi of holding liquid asses is ha in imes of low cash flow he equiy holders can survive o benefi from fuure dividends. The assumpion ha here is no accumulaion of risky asses is echnically imporan because i reduces he dimensionaliy of he dynamic programming problem o be solved for opimal dividend policy. As was menioned above we assume ha i is no possible o issue new equiy. Again his is a simple way of modelling he cosliness of exernal finance. A he cos of addiional analyical complexiy, i could be relaxed o assume equiy can be issued bu only a a posiive marginal cos ( hair cu ). Noice ha he reurn on asses is decreasing in he amoun of liquid asses held. As a consequence, he decision o increase he proporion of asses held in liquid form will reduce he growh rae of he firm. Saed in anoher way holding liquid asses inside he firm as opposed o paying hem ou and making hem available for curren consumpion reduces he value of he (cum dividend) value of he firm. Thus he firs bes rule in hese circumsances is o hold no liquid asses. This will imply ha he levered firm eners bankrupcy almos surely. (This is he consequence of he cash flow process having unbounded variaion.) However, in he absence of bankrupcy coss, his involves no loss of cash flows -- cash flow righs simply are ransferred from old equiy holders o deb holders who become he new shareholders. The policy ha maximizes he value of shares will in general differ from his firs bes policy. Finding his policy involves solving a dynamic programming problem wih wo sae variable, he curren drif of he cash flow process, ρ, and he curren level of liquid asse holdings, C. The dividend payou policy is given by ds Id dc. Noice ha we do no impose any cos of changing dividend policy from one momen o he nex. Then in his seing we can apply fundamenal resuls from he heory of conrolled Brownian moion o noe ha he opimal dividend policy will be o payou all available cash flow and liquid asses in excess of a criical amoun and o reain all available cash flows if liquid asse 7 The reurn on he firm's liquid asses may be sricly less han he risk-free rae if heir availabiliy creaes agency problems of he sor described by Jensen and Meckling (1976) or Myers and Rajan (1998). 5

6 holdings are less han his criical amoun. (See, Dua and Radner (1999)). Specifically, a basic propery of he model is: Proposiion 1: There is a criical level of cash holdings, C *( ρ ), such ha if he curren level expeced cash flow rae is ρ he firm will pay ou all available liquid asses in excess of C *( ρ ) as dividends and if liquid asses are currenly less han C *( ρ ) he firm will pay no dividends and will reain all cash flow in excess of deb service and oher fixed charges in order o accumulae addiional liquid asses. I is imporan o poin ou ha an implicaion of his proposiion is ha when cash flow ends o be high (high ρ ) here will generally be a dividend paid which will flucuae wih movemens of cash flows. In his high cash flow region he liquid asse holdings will be mainained close o he criical level C *( ρ ) and will change only as he rae of expeced cash flows ( ρ ) changes. In conras, when expeced cash flows are low (i.e., low ρ ), hen here will generally be a zero dividend and liquid asse holdings will flucuae in sep wih cash flows. In his laer region, observed liquid asse holding ( C ) may be far less han desired levels of liquid asse holding, ( C ( ρ ) ). * The soluion of he opimal policy in his conex consiss in characerizing he criical level of liquid asse holding C *( ρ ) for all possible expeced raes of cash flows. This is a complicaed dynamic programming problem, which we have solved numerically. Some imporan characerisics of he soluion can be seen in he graph where we have ploed C *( ρ ) as a funcion of ρ under specific parameric assumpions. In his graph we have represened he soluion for a low deb case, which is given by he doed line, and for a high deb case given by he solid line. In he high deb case he parameer I is 50% higher han in he low deb case. Opimal Liquid Asse Holdings Expeced Cash Flow 6

7 In examining he low deb case (he doed line) noice ha he opimal level of liquid asse holdings is a decreasing funcion of expeced cash flows over he region of relaively high expeced cash flows. The inuiion for his is ha, given ha he currenly high cash flows are likely o persis, he prospec of financial disress is relaively remoe. The firm need no carry very high levels of liquid asses. However, should expeced cash flows begin o drop he shareholders will begin o arge higher levels of liquid asse holding as a precauion agains a ime when cash flows will fall shor of conracual coss. Noe however ha he opimal liquid asse holding is an increasing funcion of expeced cash flows in he region of low expeced cash flows. The reason is ha if expeced cash flows are low relaive o coss (Id) hen he likelihood is ha available liquid asses will be drawn down during he period of firm poor performance. A some poin he shareholders decide ha prospecs of he firm are oo poor and ha i is beer o pay ou a large fracion of liquid asses as dividends. In so doing, hey willingly increase he likelihood of insolvency and herefore sacrifice fuure dividends. We can summarize his discussion in, Proposiion 2: There is a criical level of expeced cash flows, ρ * = arg max( C *( ρ )), such ha for ρ < ρ *, he criical level of liquid asses C *( ρ ) is weakly increasing in he expeced cash flow and for ρ > ρ * he criical level C ( ρ ) is weakly decreasing in he expeced cash flow. * Noice ha he non-monooniciy of he relaion beween opimal liquid asse holding and expeced cash flows suggess ha here will be no simple empirical relaionship beween observed cash flows and he level of liquid asse holding. If realized cash flow and liquid asse holdings were observed for a firm over an exended period of ime i would no be surprising o find a posiive correlaion or a negaive correlaion or no significan correlaion a all. If we compare he graphs of he criical liquid asses for low deb (doed curve) and high deb (solid curve), we noice ha he criical level of cash flows is higher for he high deb firm han for he low deb firm, ρ * L < ρ * H. Furhermore, for expeced cash flows greaer han he criical level for he high deb firm, ρ * H < ρ, he arge level of cash holdings for he high deb firm exceeds ha of he low deb firm, C * H ( ρ ) > C * L ( ρ ). If we combine his fac wih he observaion made above ha in he range of relaively high expeced cash flows acual liquid asse holdings are ypically close o arge holdings whereas for relaively low expeced cash flows acual liquid asses holdings are generally far less han arge holdings we arrive a an imporan empirical implicaion of his analysis. Everyhing else equal high levels of deb end o be associaed wih high levels of liquid asse holdings. In his secion we have limied our discussion o he simples version of our coninuous ime model of corporae liquid asse holding in order o highligh he implicaions of he model for changes in financial srucure, he level of liquid asse holdings and he reurn of asses for he firm. 8 For reasons of limiaion of space we have no presened he echnicaliies 8 I is worh poining ou how his analysis differs from hose of Kim e.al. and Morellec discussed in Secion 3. Kim e.al. model is saic in naure. The moivaion for holding liquidiy is o avoid issuing cosly deb in order o pursue growh opporuniies ha arise. There is nohing comparable o he precauionary moive ha we have emphasized. Morellec model is dynamic in naure, bu his analysis of he liquidaion decision is saic. Specifically, he allows for a one-ime asse sale. In conras, our analysis allows for liquid asses o be sold or o 7

8 involved in solving his model and we have no explored is full implicaions. For example, he model allows us o consider he effec of increasing cash flow volailiy, σ, he mean cash flow rae, ρ, or he volailiy of he expeced cash flow rae, η. Furhermore, wih a sligh modificaion o allow for deb relaed corporae ax shields, he framework could be used o derive he opimal capial srucure. 4. Empirical ess 4.1 Hypoheses o be esed. The analysis of he preceding secion has served o demonsrae a link beween capial srucure and firm liquidiy, which we believe has been overlooked in previous discussions of he deerminans of firm liquidiy. I is his prediced posiive relaionship beween leverage and liquidiy ha we will focus on in our empirical analysis. However, we recognize ha leverage may be only one of he facors ha deermine a firm s choice of liquid asse holdings. Some oher facors are suggesed by oher facors ha appear explicily in our model. In paricular, we would expec asse volailiy o be posiively associaed wih liquid asse holding everyhing else equal. There are very likely oher imporan facors affecing he choice of liquid asses even hough i has no been possible o incorporae hem ino our analysis. Foremos among hese is he presence of growh opporuniies which we would expec o be posiively associaed wih liquid asse holding. The esing mehodology we adop should be careful o conrol for hese addiional affecs appropriaely. In our review of he lieraure on he firm s choice of liquid asse holdings, we have discussed he wo recen empirical sudies ha are closes o our sudy (Kim, Mauer and Sherman (1998) and Opler, Pinkowiz, Sulz, and Williamson (1999)). I is useful o summarize he findings of hese sudies and o compare hem o our predicions based on he model we have summarized in Secion 3 and he ou-of-model predicions ha we find unexcepionable. The following able does his: Table 1: Facors deermining liquid asse holdings Variable Kim e.al.finding Opler e.al. finding Our predicion Cash flow - + Ambiguous Growh opporuniies Volailiy Leverage Bankrupcy risk - n.a. Ambiguous Firm size n.a. - n.a. Noice ha he Kim e.al. sudy finds a negaive effec of cash flow on liquid asse holding and ha Opler e.al. find a posiive relaion. This apparen puzzle is in some sense resolved be accumulaed freely a all imes. This is a more complicaed problem which resuls in he fac ha we need o solve i numerically. 8

9 by our model which poins ou explicily ha he relaion beween cash flow (undersood as a proxy for expeced cash flow) and liquidiy is non-monoonic. The previous sudies agree in finding posiive relaions wih respec o growh opporuniies and volailiy and are consisen wih our predicions. The major difference beween he previous sudies and our predicions concerns he effec of leverage. Boh previous sudies found a negaive effec; whereas, our predicion is of a posiive effec. I should be noed ha leverage was included in he earlier sudies as a conrol and he auhors provided no deailed raionale for he negaive sign obained. Furhermore, he earlier sudies employed measures of leverage ha aggregaed deb of all mauriies. In U.S. corporae financial srucures shor-erm deb ends o predominae. In conras, he formal model we described in Secion 3 was based on he assumpion ha deb was long-erm (indeed perpeual). As a resul, we should be aler o he fac ha he design of he previous empirical sudies may no have been suied o esing our predicions. We will ry o deal wih his by including measures of boh long-erm and shor-erm deb in our analysis. Finally, we should poin ou ha one of he puzzling findings of he Kim e.al. sudy was ha bankrupcy risk had a negaive effec on liquid asse holdings. The measure of bankrupcy risk hey employed was based on he Alman z-score which includes a variey of financial raios ha are posiively correlaed wih cash-flow. Therefore, we noe ha he nonmonooniciy resul of our model suggess ha his relaionship may no necessarily be a robus or sable one. 4.2 Tess based on a panel of UK firms Our firs empirical esimaes are based on an unbalanced panel daa se of lised UK firms. The focus on lised firms is moivaed by several consideraions. Firs, hese firms end o be larger firms which would end o have access o long-erm deb finance. Second, hese firms would end o be hose for which more reliable esimaes of R&D expendiures (one of our proxies for growh opions) are available. Finally, for hese lised firms i is possible o calculae he raio of marke value o book value which may be an alernaive indicaor of growh opions. The daa se is drawn from Daasream consising of financial variables of non-financial firms included in he FTSE All Shares index in Our daa is covers he annual repors over a maximum of 12 years per firm, 1989 hrough Using he FTSE all shares index means ha we cover large and medium sized firms based in he U.K. The fac ha we have used firms ha were in exisence in 2001 creaes a possible survivorship bias. Some of he firms ha have disappeared from he sample may have done so hrough mergers. I is someimes saed ha highly liquid firms are aracive akeover arges. Our daa will no allow us o idenify his effec. Oher firms may have disappeared hrough bankrupcy or disressed reorganizaion. Our model suggess ha he choice of liquid asse holding by disressed firms (low ρ ) will be very differen from ha of firms under normal condiions. To he exen ha he consrucion of our daa se excludes 9

10 disressed firms, his means ha our resuls are beer characerizaions of firms under normal operaing condiions. Table 2 summarizes he definiions of he variables used in our sudy of UK firms. The definiions we have adoped for liquid asses, cash flow, R&D, and marke o book are direcly comparable o he definiions used by Kim e.al. and Opler e.al. As in heir sudies he dependen variable of our regression analysis is liqura, he oal liquid asse holding of he firm expressed as a fracion of oal asses. Noe ha all variables have been divided by a scale indicaor in order o reduce problems of heeroscedasiciy. Table 2: Daa definiions liqura Liquid asse Sum of cash, bank balances, and invesmens in curren asses divided by oal asses cfraio Cash flow Earnings before axes and ineres divided by oal asses llev Long erm deb Deb greaer han 5 years o mauriy divided by oal asses mlev Medium erm deb Deb greaer han 1 year and less han 5 years o mauriy divided by oal asses slev Shor erm deb Deb payable wihin 1 year divided by oal asses rdraio R&D Expendiures on R&D divided by oal sales expendiures MTBV Marke value o book value Marke value divided by ne angible asses, marke value is share price muliplied by he number of ordinary shares ousanding; ne angible asses is ordinary shareholder's equiy less inangible asses. R&D expendiures and marke o book raio are included as a proxies for he firm s growh opporuniies. In our view R&D is he preferred variable in ha he link o growh appears o be he cleares. Is drawback is ha no all firms repor R&D expendiures. As a resul, we have included marke o book raio which can be calculaed for a larger se of firms. This variable is no ideal because high marke o book may be indicaive condiions favorable o he firm ha have nohing o do wih is fuure growh opporuniies. Unlike previous sudies we disinguish long-erm and shor-erm deb. UK financial repors are unusual in ha hey include indicaors of debs wih a leas five years o mauriy. Thus o clearly focus on he predicions of our model concerning deb as a persisen feaure of he firm s capial srucure we ake long-erm deb o be a leas five years and shor erm years o be deb of a maximum mauriy of one year. As an addiional conrol we do some esimaions including medium erm deb (beween one and five years). In examining summary saisics for he variables included in our analysis i was noed ha some implausibly high or low values of variable were observed. These may well reflec daa errors. We deal wih his by working wih a runcaed sample where he daa poins involving he smalles percenile or larges percenile of a variable are excluded. In addiion o he variables we have lised we have included in he analysis dummy variables for indusrial caegories and year of observaion. We have also explored he effec of alernaive definiions of leverage (including rade credi), lagged values of explanaory variables, and measures of cash flow volailiy. I is no possible o presen all hese 10

11 esimaes here. Generally, we have found ha indusry dummies and year dummies are significan. Furhermore, once hese indusry dummies are included he esimaed effec of volailiy is no significan; alhough, his variable is significan when indusry dummies are excluded. The inclusion of lagged regressors did no produce any sysemaic or ineresing dynamics. The sign and significance of hese variables were sensiive o exac specificaion and esimaion mehod. There inclusion had lile effec on he sign and significance of he conemporaneous regressors. Therefore we repor he resuls excluding hese lagged regressors. We have esimaed he linear regression model beween liquidiy raio and he explanaory variables we have lised. We have esimaed his relaion by Ordinary Leas Squares. Furhermore, given he naure of our daa se i is ineresing o esimae he model using panel sudy echniques. A limiaion of hese echniques is ha ofen he lengh of he panel is ofen oo shor o yield powerful resuls. In our case, we have a maximum of 11 years of daa for a given firm. We consider his sufficien o make hese echniques poenially ineresing. We repor resuls for boh firm fixed effecs and random effecs versions of our models. 4.3 UK Resuls Table 3 repors he resuls of he ordinary leas squares regressions of he model based on he sample of UK firms described in he preceding secion. Columns 1 presens he esimaes including indicaors for long-erm and shor-erm leverage and using R&D expendiures as a proxy for growh opporuniies. Since R&D is our preferred proxy for growh we view his regression as a benchmark for comparison wih oher resuls. Cash flow eners wih posiive sign and is significan. Table 3: UK Firms, rimmed sample, OLS regressions, dependen variable liquidiy raio (-saisics under parameer esimaes) cfraio llev mlev slev rdraio MTBV year dummies yes yes yes yes indusry dummies yes yes yes yes 11

12 Adjused R Number of observaion s Long erm leverage has a posiive coefficien which is highly significan. This is consisen wih he predicion of our heoreical model. Shor-erm leverage eners wih a negaive sign bu is insignifican. R&D spending is posiive and highly significan. The indusry dummies aken as a whole are significan. The same applies o he year dummies aken as a whole. Column 2 differs from column 1 in ha marke o book raio is aken as a proxy for growh opporuniies. This regression is esimaed on a larger daa se since a large number of firms in he UK sample do no repor expendiures on R&D spending. The resuls are consisen wih he esimaes of column 1. The proxy for growh opporuniies eners wih posiive sign and is significan. The oher variables ener wih he same signs as in column 1. However, cash flow and shor-erm leverage are insignifican and long-erm leverage is only marginally significan. Columns 3 and 4 differ from columns 1 and 2 in ha hey include medium leverage (based on deb wih 1 o 5 years mauriy) in he regression. In boh cases his variable eners wih negaive sign and is significan. Oherwise he paern of esimaes is similar o ha seen in columns 1 and 2. Long-erm leverage has a posiive coefficien and is highly significan in he regression where R&D is a proxy for growh opporuniies. When marke o book is used o conrol for growh opporuniies his variable sill has a posiive effec, bu i is no longer significan. In boh columns 3 and 4 shor-erm leverage eners wih negaive sign bu is insignifican. Growh opporuniies (R&D or marke o book) ener posiively and are significan. The major finding from Table 3 is ha oher hings equal, greaer long-erm deb is associaed firms seing higher arge levels of liquid asses. This is consisen wih he heory of he choice of liquid asses we have elaboraed above. Furhermore, hese resuls raise doubs abou he finding of previous sudies which a negaive relaionship beween firm liquidiy and firm indebedness. In our daa se shor-erm leverage enered negaively ino he relaion and was insignifican. Medium erm leverage enered negaively and was significan. These resuls sugges ha firms may view long-erm deb and shor-erm deb very differenly. Long-erm deb may reflec a durable feaure of he firm s capial srucure o which oher policies, such as dividend and liquid asse holding are adaped. Shor-erm deb can be used more acively and may be subsiuable for liquid asse holding. In he regression analysis repored in Table 3 here is a risk ha here are deerminans of liquid asse choice beyond he measures of leverage, cash flow, and growh opporuniies ha we have explicily inroduced. We have aemped o conrol oher possible facors by including dummy variables for ime and he indusry of he firm. These may be imperfec conrols. The panel srucure of our daa se allows us o ake an alernaive approach by allowing for firm specific effecs which are eiher consan over ime (fixed effecs) or are drawn randomly each year from a fixed disribuion (random effecs). In Table 4 we repor he resul of applying hese panel echniques o our baseline model specificaions which include leverage measures, cash flow, growh opion proxies, and dummies for indusry and year. Esimaion is by generalized leas squares. 12

13 Table 4: UK firms, panel esimaes, rimmed sample, dependen variable liquidiy raio (z-saisics or -saisics under parameer esimaes) Random effecs Fixed firm effecs Random effecs Fixed firm effecs Cfraio Llev Slev Rdraio MTBV year dummies yes yes yes yes Indusry dummies yes yes yes yes Haussman es (p-value) Number of obs Columns 1 and 2 of Table 4 presens he panel esimaes based on R&D spending as a proxy for growh opporuniies. The resuls of he Haussman es sugges ha he fixed effecs regression is he preferred specificaion; however, he coefficien esimaes are similar in he wo specificaions. Long-erm leverage eners wih a posiive sign and is significan. Ineresingly, shor-erm leverage eners negaively and is significan in boh specificaions. Cash flow eners posiively. I is significan in he random effecs version of he model bu no in he fixed effecs version. As expeced R&D eners posiively and is highly significan in boh he random effecs and fixed effecs versions of he model. is no longer significan. Shor-erm leverage has a negaive coefficien and is significan. The marke o book raio eners posiively bu is insignifican. This ends o confirm our reservaions expressed abou his measure as a proxy for growh opporuniies. Finally, cash flow eners posiively in he random effecs version of he model bu negaively in he fixed effecs version. In boh cases i is insignifican. The resuls of hese panel esimaes end o confirm he main conclusions enaively reached on he basis of he OLS regressions. We find significan evidence of a posiive relaionship beween long-erm leverage and liquid asse holding. This is consisen wih he presence of a precauionary moive for holding liquid asses for firms ha mainain high leverage as a durable feaure of heir capial srucure. Beyond his here seems o be evidence of a 13

14 negaive relaionship beween shor-erm leverage and asse liquidiy. This suggess ha shor erm deb and liquid asse holding may be subsiues in he sense ha a firm facing persisenly low cash flows will respond eiher by drawing down available liquid asses or by accumulaing shor erm deb or boh. Anoher poin ha emerges from our analysis of he panel daa se of UK firms is ha he relaionship beween cash flow and liquid asse holding does no appear o be sable or robus. In he resuls repored here as well as in our experimens wih alernaive specificaions of he model, we found ha he esimae of he cash flow were someime posiive, someimes negaive and ofen insignifican. Our heoreical analysis of Secion 3 suggess ha his should no be surprising. 4.4 Tesing based on a panel of Belgian firms To furher sudy he deerminans of he firm s choice of liquid asse holding we have exended our analysis o an unbalanced panel daa se of lised Belgian firms. This is based on he annual financial repors of Belgian firms for 1986 hrough 1999 as compiled by he Naional Bank of Belgium. From his daa se we have seleced non-financial firms ha have been lised on he Belgian sock exchange. As in he case of he UK daa, he focus on lised firms mean ha we have sudied relaively large firms which had access o capial markes and which migh provide daa on R&D spending (our main proxy for growh opporuniies). I should be noed ha by excluding non-financial firms we have excluded many of he larges Belgian firms which collecively have represened a large share of Belgium s oal marke capializaion. In order o aid comparabiliy of resuls we have ried o work wih variables ha are as close as possible o hose we have used in our UK daa base. In mos cases he correspondence is close, bu several differences should be noed. Mos imporanly, Belgian financial repors define shor-erm deb as payable wihin one year and all oher deb considered long-erm. Thus here is no measure direcly comparable ha for he UK which indicaed debs of greaer han five years. In comparing Belgian and Briish measure of firm deb s we are alered o possible ambiguiies in wha is mean by long-erm and shor-erm deb. For example, if a firm has negoiaed a revolving credi faciliy, by drawing upon he faciliy i may creae a shor-erm liabiliy even hough mainaining some relaively high level of his liabiliy may be a permanen feaure of is capial srucure. Anoher feaure of he NBB daa se is ha i does no include measures of he marke value of he equiy. Therefore, we have been unable o use marke o book raio as a proxy for growh opporuniies. Given our concepual reservaions abou his measure and he raher unsable resuls obained for i in he UK daa se, we do no consider his a major limiaion of our sudy of Belgian firms. As wih our sudy of UK firms esimae he linear regression model of he deerminans of he raio of liquid asses o oal asses. Esimaes are obained boh by he mehod of ordinary leas squares and by panel mehods. In sudying he descripive saisics of our variables we have found evidence of ouliers, which migh be aribuable o reporing or 14

15 coding errors. All he resuls we repor are based on he rimmed sample wih hese apparen ouliers eliminaed. 4.5 Belgian resuls Before aemping o explain he deerminan of liquidiy, i is useful o poin ou an imporan apparen difference beween our UK and Belgian daa ses. Table 5 repors he quariles of he disribuion of liquidiy raio for hese wo daa ses. Based on hese daa we see ha he level of liquid asse holding of Belgian lised non-financial companies appears o be considerably higher han ha of heir UK counerpars. For example, he median liquid asse holding is 11% in Belgium versus 7% in he UK. The high end of he disribuion is quie surprising. Abou 25% of Belgian non-financial firms hold a leas 23% of heir asses in liquid form. In conras, he op quarile of UK firms holds a leas 14% of heir asses in liquid form. These observaions are sufficien o poin ou ha liquid asse holding even for indusrial and commercial firms can be quie subsanial. Table 5: Comparisons of corporae liquidiy (raio of liquid asses o oal asses) 25 h percenile median 75 h percenile UK Belgium The resuls of esimaing he linear model of he deerminans of liquid asse holding are repored in Table 6. Column 1 is devoed o variables included in he regressions are he same for each column. The columns differ wih respec o esimaion mehod. Oherwise he specificaions of he models are idenical. We find a posiive and significan relaionship beween he level of liquid asse holding and growh opporuniies as proxied by R&D spending. We also find a posiive and significan relaionship wih respec o shor erm leverage. In conras, we find ha long-erm leverage eners negaively and is insignifican once firm effecs (eiher fixed or random) are inroduced o he regression. Cash flow raio eners negaively in he OLS esimaes bu posiively in he panel esimaes. 15

16 Table 6: Belgian firms, Trimmed-sample regressions, dependen variable liquidiy raio (-saisics or z-saisics under parameer esimaes) OLS Fixed effecs Random Effecs Cfraio Llev Slev Rdraio year dummies yes yes yes indusry dummies yes yes yes Adjused R Number of observaions Haussman ex (p-value) The main apparen difference beween he resuls we have obained for Belgian firms and hose we obained UK daa se is he reversal of sign paerns for long-erm and shor-erm leverage. There is no obvious or simple accouning for his. One possible source of discrepancy ha we have already noed is ha long-erm liabiliies coming due in one year or more for Belgian daa whereas in he UK daa se i perains o debs greaer han 5 years. However, i is no clear why his difference, in iself, explains he fac ha shor-erm deb is posiively associaed wih liquid asse holding in Belgium; whereas, in he UK i is long-erm deb ha is posiively associaed wih leverage. Raher, i seems ha our resuls poin o some oher insiuional differences beween he Belgian and UK conexs ha are masked by he sandard accouning definiions. We have already menioned he poenial problem ha a long-erm revolving credi faciliy migh be recorded as shor-erm deb even hough i is a permanen feaure of he firm s capial srucure. Thus if his kind of arrangemen more prevalen in Belgium as compared o he UK we migh beer undersand why in Belgium i appears ha shor-erm deb creaes a srong precauionary moive for liquid asse holding. Unforunaely, our daa does no allow us o deermine wheher his conjecure is correc. Therefore, his issue mus awai furher sudy for clarificaion. Finally, we noe ha in Table 6 he cash flow variable eners negaively in he OLS regression. In he panel esimaes he sign of his variable becomes posiive. This apparen 16

17 sensiiviy of his variable o changes of specificaion or esimaion mehod was observed wih UK daa and has been given an inerpreaion in our analyical discussion of Secion Implicaions I is useful o summarize he main sylised facs abou corporae liquid asse holdings ha emerge from he analysis of he preceding secions. Firs, wihin he universe of nonfinancial firms here are subsanial differences across firms and across counries in he proporion of asses ha are held in liquid form. Second, here appears o be a srong and robus posiive relaionship beween he presence of growh opporuniies and corporae liquidiy. Third, cash flow volailiy appears o be posiively associaed wih liquid asse holding. Fourh, here does no appear o be a sable or robus relaionship beween cash flow and corporae liquidiy. Fifh, US sudies a negaive relaion beween leverage and corporae liquidiy; whereas, our sudies of Briish and Belgian firms found evidence of a posiive relaionship beween leverage and corporae liquidiy. This las observaion highlighs an issue ha deserves furher careful sudy. Our analysis of Belgian and UK firms shows ha for he purposes of undersanding corporae liquidiy i is very imporan o differeniae deb according o mauriy. The US sudies used measures of leverage which combined debs of all mauriies. Thus, i would be very ineresing o reconsider he US daa sudied by Kim e.al. and Opler e.al. allowing for long-erm and shor-erm leverage. Furhermore, i would be ineresing o sudy he implicaions for corporae liquidiy of finer indicaors of capial srucure, e.g., bank versus oher deb, use of callable deb, or converibles. Thus our undersanding of he way capial srucure impacs corporae holding of liquid asses is sill incomplee. Neverheless, we believe ha analysis in his paper clearly raises he possibiliy of an imporan precauionary moive for corporae liquidiy in environmens where access o exernal capial is cosly. In paricular, we have idenified a channel ha operaes as follows: Liquidiy grans a survival opion o he shareholders of he levered firm. Consequenly, hese shareholders will choose a higher level of asse liquidiy han would maximize he value of he firm. In so doing, hey reduce he rae of reurn on asses and he growh of he firm. This effec is more pronounced he greaer is he leverage used by he firm. Furhermore, i is more pronounced he more rigid is he financial srucure and he more cosly is accessing exernal financial markes. In his secion we draw ou some of he possible implicaions of his channel. In his regard our analysis adds o he work on inernaional comparisons of financial developmen as a deerminan of differences in economic growh raes. In ha lieraure, he sudy ha mos clearly links differences in corporae growh o he insiuions of he financial secor is ha of Rajan and Zingales (1999). Tha sudy did no specifically explore he channel ha we have idenified here. However, implicily heir earlier inernaional comparison capial srucure (Rajan and Zingales (1995)) did produce some resuls ha are suggesive ha his channel may be imporan. In paricular, in ha sudy hey found when leverage is measured as non-equiy liabiliies divided by oal asses, he Anglo-American economies have considerably lower median leverage in 1991 (abou 0.56) han companies in Coninenal Europe and Japan (0.70). (Rajan and Zingales (1995), p. 1429). However, when hey 17

18 deduc liquid asse holding from non-equiy liabiliies, hey find ha he leverage levels are similar across he counries. If anyhing Germany emerges as a low level counry. By implicaion high leverage is associaed wih high levels of liquid asses. We sill do no have a complee undersanding of wha accouns for differences in leverage or oher aspecs of financial srucure across counries. The sudy of Rajan and Zingales (1995) sugges ha he radiional break-down beween bank-based and marke-based sysems is a bes only a crude explanaion of observed differences. I is likely ha differences ax incenives and ownership paerns (e.g. family and inerlinked coaliions of invesors versus dispersed ownership) accoun for more of hese differences. Our highlighing he presence of a srucure-liquidiy-growh linkage brings an addiional dimension o observaions abou insiuional differences across financial sysems. Regarding he difference beween bank-based versus marke-based sysems, our analysis suggess ha a crucial aspec of financial srucure will be he flexibiliy ha i affords he firm in dealing wih periods of cash flow shor-fall. If close relaionships o crediors mean ha firms will be able o access addiional credi during hese periods, hen his sofens he precauionary moive for corporae liquidiy. While his may be a poenial advanage of sysems based on close banking relaionships, he recen work by Edwards and Fischer on German financial sysems raises doubs abou wheher he haus-bank relaionship really affords his flexibiliy. Anoher imporan observaion ha has emerged from inernaional comparisons of corporae financial pracices is ha some sysems are characerised by srucures ha help o perpeuae he effecive conrol over firms by family groups or oher coaliions of invesors. This appears o be wha accouns for he prevalence of pyramids and inerlinked ownership srucures in some coninenal European sysems. Our analysis poins ou a hidden implicaion of he reliance on deb finance in order o concenrae conrol righs over he firm in he hands of specific invesors. If invesor groups aemp reain conrol a firm hrough he heavy reliance on deb, hey may be forced o mainaining a higher degree of asse liquidiy han hey would oherwise. In so doing hey may be undermining he subsequen growh raes of he firm and condemning he firm o ulimae sagnaion. Our analysis also pus he widely observed growh of European equiy markes in a new ligh. If equiy marke developmen has effecively brough down he barriers o equiy finance for a wider range of firms, his poenially could have an imporan and long-lasing impac on he growh poenial of hese firms. If his source of finance effecively reduces he observed leverage levels, i also reduces he precauionary incenive for corporae liquidiy. Firms may be freer in channelling heir available cash flow ino riskier posiive NPV invesmen opporuniies. Furhermore, our analysis poins ou one of he favourable consequences of equiy based finance is ha i can be growh promoing even if new equiy issuance is no he direc source of finance for risky projecs. By simply creaing a srucure whereby reenions can be channelled ino risky projecs raher han low reurn liquid asses, firm growh is promoed. While our empirical analysis has focused on lised firms for reasons of daa reliabiliy, he capial srucure-liquidiy-growh channel we have idenified can be a leas as imporan for smaller firms wih srong growh poenial. Our analysis suggess ha he environmen may be seriously growh inhibiing if he only effecive means of financing growh a early sages of he firm is hrough deb. For wih increased leverage brings a sronger liquidiy bias hus 18