GROWTH STRATEGIES AND CAPITAL STRUCTURES OF SMALL AND MEDIUM-SIZED ENTERPRISES *

Transcription

1 GROWTH STRATEGIES AND CAPITAL STRUCTURES OF SMALL AND MEDIUM-SIZED ENTERPRISES * Mnna Martkanen a Juss Nkknen b a Lappeenranta Unversty of Technology, Fnland b Unversty of Vaasa, Fnland February 5, 2007 Abstract We examne the mpact of growth strateges on captal structures usng a random sample of 1,153 non-lsted ncorporated small and medum-szed enterprses (SMEs). We dentfy the frms wth the followng four growth strateges: (1) new product development, (2) ncreasng export, (3) networkng, and (4) marketng and sales promoton. Our results show that frms wth growth strateges have more long-term and short-term debt than comparable frms wth no growth strateges. We also fnd that the type of the growth strategy affects the choce between long-term and short-term debt. Our results reveal that networkng and n partcular marketng and sales promoton ncrease the use of short-term debt, whereas new product development and export actvtes requre more long-term fnancng. Furthermore, a frm s captal structure s found to depend on a frm s ndustry, proftablty, rsk, realzed growth, asset structure, and frm sze and age. JEL classfcaton: G30; G32; G35 Keywords: captal structure, debt ratos, growth strategy, small and medum-szed enterprses * We are grateful to the Federaton of Fnnsh Enterprses for the survey data and to Statstcs Fnland for provdng the fnancal statement data and research facltes. a Mnna Martkanen, Lappeenranta Unversty of Technology, Department of Accountng and Fnance, P.O. Box 20, FIN Lappeenranta, Fnland; Tel ; E-mal address: mnna.martkanen@lut.f b Correspondng author: Juss Nkknen, Unversty of Vaasa, Department of Accountng and Fnance, P.O. Box 700, FIN Vaasa, Fnland; Tel ; Fax: ; E-mal address: jn@uwasa.f

2 1 GROWTH STRATEGIES AND CAPITAL STRUCTURES OF SMALL AND MEDIUM-SIZED ENTERPRISES Abstract We examne the mpact of growth strateges on captal structures usng a random sample of 1,153 non-lsted ncorporated small and medum-szed enterprses (SMEs). We dentfy the frms wth the followng four growth strateges: (1) new product development, (2) ncreasng export, (3) networkng, and (4) marketng and sales promoton. Our results show that frms wth growth strateges have more long-term and short-term debt than comparable frms wth no growth strateges. We also fnd that the type of the growth strategy affects the choce between long-term and short-term debt. Our results reveal that networkng and n partcular marketng and sales promoton ncrease the use of short-term debt, whereas new product development and export actvtes requre more long-term fnancng. Furthermore, a frm s captal structure s found to depend on a frm s ndustry, proftablty, rsk, realzed growth, asset structure, and frm sze and age. JEL classfcaton: G30; G32; G35 Keywords: captal structure, growth strategy, small and medum szed enterprses

3 2 1. Introducton Small and medum-szed enterprses (SMEs) represent substantal proportons of Western economes. For example, Dens (2004) among others notes that although large frms have hstorcally been job creators, the stuaton has reversed n the last twenty years, as the number of SMEs and jobs created by them has remarkably ncreased. Gven the ncreased mportance of SMEs, surprsngly few studes have focused on the entrepreneural accountng and fnance ssues. 1 Exceptons nclude Ang (1991), Holmes and Kent (1991), Berger and Udell (1995), Esenberg, Sundgren and Wells (1998), Ang, Cole Ln (2000), Cavalluzzo, Cavalluzzo and Wolken (2002), Cassar and Holmes (2003), and Hall, Hutchnson and Mchaelas (2000, 2004). In ths study, we follow a research avenue of Holmes and Kent (1991), Hall et al. (2000, 2004) and Cassar and Holmes (2003) among others, whch focuses on the captal structures n SMEs. The specfc objectve of our study s to examne the mpact of growth strateges on captal structures. Investgatng the effects of growth strateges usng SME data s partcularly well-motvated snce, for nstance, Cassar and Holmes (2003) pont out that many smaller scale frms may have strong growth opportuntes. Furthermore, for example, Harrs and Ravv (1991) state: Ths area [models whch related captal structure to products and nputs] s stll n ts nfancy and s short on mplcatons relatng captal structure to strategc varables, etc. Along these lnes, Cassar and Holmes (2003) also suggest that frm strategy could be a potentally useful varable n explanng captal structure. The captal structures of large lsted frms are usually explaned by usng two theores of Myers (1984) and Myers and Majluf (1984), the tradeoff theory and the peckng order 1 For example, Dens (2004) notes that there has also been a dramatc ncrease n the amount of captal allocated to prvate frms, but at the same tme research on entrepreneural fnance has lagged behnd ths development.

4 3 theory, whch the exstng emprcal evdence has been generally found to support (see, e.g., Harrs and Ravv 1991, Rajan and Zngales 1995, Wald 1999, Korajczyk and Levy 2003, and Brounen, de Jong and Koedjk 2006). As ponted out by Ang (1991) and Hall et al. (2004), the captal structure ssue for the small frm, however, may be somewhat dfferent from ther larger counterpart. Accordng to Ang (1991) and Berger and Udell (1995), the man dfferences are related to sources of funds and barganng between owners and these sources of fnancng, but t may be noted that there s also a peckng order for the funds. For example, a hgh preference s to use the funds generated wthn the busness and as also ponted out by Hall et al. (2000), a further equty nvestment by owner-managers would rank ahead of borrowng. Emprcal evdence usng SME data seems to reflect these dfferences between small and large frms. For example, Holmes and Kent (1991) compare the fnancal structure of small and large Australan manufacturng frms and fnd that because of lmted ablty to obtan external equty fundng the fnancal structures of small and large frms are sgnfcantly dfferent. Hall et al. (2000) and Cassar and Holmes (2003) examne the longand short-term debt patters of unquoted, U.K. and Australan SMEs. These studes and Hall et al. (2004) report a postve relaton between long-term debt and the asset structure of a frm. Hall et al. (2000, 2004) moreover document that long-term debt s postvely related to frm sze and negatvely related to frm age and Cassar and Holmes (2003) fnd that longterm debt s negatvely related to proftablty. Short-term debt n turn s found to be negatvely related to proftablty and asset structure n all these studes. Furthermore, Hall et al. (2004) extend the prevous analyses by examnng the degree to whch the determnants of SMEs' captal structures dffer between European countres and document the country and frm-specfc effects. In sum, the studes on SMEs cted above are broadly consstent wth

5 4 the peckng order theory and the fndngs also appear to be broadly consstent across dfferent ndustres, countres and busness envronments. 2 In ths paper, we contrbute to the exstng lterature by examnng the relaton between growth strateges and captal structures. We frst dentfy frms wth the followng four growth strateges: (1) new product development, (2) ncreasng export, (3) networkng, and (4) marketng and sales promoton. We then nvestgate, as our prmary research task, the effects of growth strateges on the long and short-term debt to total assets ratos. Motvated by the arguments presented n Ang (1991) and Cassar and Holmes (2003), we hypothesze that when non-lsted SMEs mplement growth strateges, the process requres external debt fnancng. We moreover hypothesze that the type of a growth strategy affects the maturty structure of debt. For example, producton development and exportng may be expected to requre long-term fnancng. By contrast, marketng and sales promoton n turn may take place n campagns of lmted length, for whch short-term fnancng may be more approprate. We examne the relatonshp between growth strateges and captal structures usng a random sample of 1,153 Fnnsh non-lsted ncorporated SMEs. Fnland provdes an nterestng settng for ths type of examnaton snce t consttutes an deal example of a compettve and nnovatve busness envronment. 3 To llustrate the mportance of SMEs n Fnland, they were estmated to represent 99.7 % of all Fnnsh busness enterprses and to employ 61 % of the workforce n the prvate sector at the end of The comprehensve fnancal statement data from Statstcs Fnland used n our study were ntally compled by 2 Hall et al. (2004) conclude that the hypotheses appear to hold up reasonably well, but they also note that the hypotheses do not explan everythng n terms of SME captal structure snce there are varatons n the effects of the determnants on captal structure between countres. 3 Fnland has frequently been ranked as one of the world s most compettve busness envronments. See, for example, the Global Compettveness Report by World Economc Forum, and the World Compettveness Yearbook 2006 by IMD.

6 5 the tax authortes n Fnland. Smlar type of fnancal statement data on Fnnsh small frms has prevously been used, for example, by Esenberg et al. (1998). Accordng to our results, networkng, and marketng and sales promoton strateges are typcal for younger frms, and as s ncreasng export for older frms. The use of growth strateges typcally ncreases wth frm age. Our results show that growth orented frms wth growth strateges have more both long-term and short-term debt than comparable frms wth no growth strateges. We also fnd that the type of the growth strategy affects the use of long and short-term debt. Our results suggest that networkng and n partcular marketng and sales promoton ncrease the amount of short-term debt, whereas new product development and export actvtes seem to requre more long-term fnancng. Consstently wth Hall et al. (2000, 2004) and Cassar and Holmes (2003), a frm s captal structure s found to depend on the ndustry, proftablty, rsk, growth, asset structure, and frm sze and age. The remander of ths study s organzed as follows. Secton 2 presents the fnancal statement and growth strategy data used n the analyss. Secton 3 reports the emprcal fndngs of the effects of the growth strateges on the long and short-term debt ratos of SMEs. It also presents a user profle for each growth strategy. Secton 4 concludes.

7 6 2. Fnancal statement data and prelmnary data analyss The fnancal statement data on the non-lsted Fnnsh frms used n ths study were obtaned from Statstcs Fnland. The database contans the fnancal statements of Fnnsh frms, and s based on the busness taxaton regster (EVR) for whch the datawere compled by the tax authortes ntally for taxaton purposes. We focus on corporatons excludng ndvduals, partnershps, etc. from the sample. 4 It may be noted that the fnancal statements of these frms have been audted. As an addtonal data screenng crteron, the frms are requred to employ at least 3 persons and to be at least 7 years old. We nclude the frms from the sx largest man ndustres accordng to SIC classfcaton provded by Statstcs Fnland. 5 The ndustres are (D) manufacturng, (F) constructon, (G) wholesale and retal trade; repar of motor vehcles, motorcycles and personal and household goods, (H) hotels and restaurants, (I) transport, storage and communcaton, and (K) real estate, rentng and busness actvtes. Gven the above screenng crtera, our random sample ncludes 1,153 SMEs. In the followng two sectons, leverage and growth strategy measures and control varables are descrbed Leverage and growth strategy measures Followng Hall et al. (2000, 2004), and Cassar and Holmes (2003) among others, we measure the captal structure usng separately the rato of long-term debt to the book value of total assets (LTD) and the rato of short-term debt to total assets (STD) n the analyss. These long-term and short-term leverage ratos are computed usng the fnancal statements from 4 Ths data screenng s smlar, for example, to Esenberg et al. (1998). 5 Fnland's natonal standard ndustral classfcaton s based on the European Unon classfcaton of economc actvtes NACE 2002.

8 Includng the long-term and short-term leverage ratos separately allows an examnaton of the effects of growth strateges on the maturty structure of debt rather than the total debt poston of sample frms. As the source of growth strateges we use a jont survey of the Federaton of Fnnsh Enterprses and Fnnvera, Ltd., whch provdes nformaton on growth strateges of sample frms. To be precse, the survey ndcates whether the frm has a growth strategy and specfes ts type. We dentfy the frms wth the followng four growth strateges: (1) new product development, (2) ncreasng export, (3) networkng, and (4) marketng and sales promoton. As the next step, we use a bnary varable approach to dentfy frms wth growth strateges. Consequently, we use a dummy varable to ndcate frms wth growth strateges and four separate dummy varables to dfferentate between the growth strateges. We hypothesze that when non-lsted SMEs mplement growth strateges, the mplementaton usually requres external debt fnancng. Ths expectaton s consstent wth suggestons n Ang (1991) and Cassar and Holmes (2003). For example, Cassar and Holmes (2003) suggest that frm strategy could be a potentally useful varable n explanng captal structure. We moreover expect that the type of growth strategy affects the maturty structure of debt. For example, producton development and exportng may be expected to requre long-term fnancng. By contrast, marketng and sales promoton may take place n campagns, for whch short-term fnancng may be more approprate. Snce there exst several frm specfc characterstcs and ndustry effects whch may affect the debt usage, several control varables are ncluded n the analyss Control varables

9 8 A set of control varables s used n the multvarate analyss to take nto account ndustry effects (see, e.g. Hall et al. 2000) and varous frm characterstcs (see Wald 1999, Hall et al. 2000, Cassar and Holmes 2003, and Hall et al. 2004). Followng Wald (1999), Hall et al. (2000, 2004) and Cassar and Holmes (2003) asset structure can be hypotheszed to be a determnant of a frm s captal structure. As n Wald (1999), we nclude separately the rato of property, plant, and equpment to the book value of assets (PP&E) and the rato of nventores to the assets (INV). Our data clearly supports ths specfcaton over the alternatves. 6 However, we also estmated the models as, for example, n Cassar and Holmes (2003), and Hall et al. (2004) to check that our prmary results are not senstve to the selected specfcaton. As n earler studes, we may expect that there exsts a postve relaton between asset structure and long-term debt (LTD) and a negatve relaton between asset structure and short-term debt (STD). Furthermore, followng Wald (1999), we nclude the rato of deprecaton to total assets (DEPR) as an addtonal control varable. We use the standard devaton of the frst dfferences n return on assets as the rsk measure (RISK). We use a fve-year estmaton perod. Our choce s therefore smlar to Wald (1999). As shown by Cassar and Holmes (2003), the prmary results are not senstve to the choce or excluson of the rsk measure. Based on the earler lterature such as Wald (1999) t may be expected that there exsts a negatve relatonshp between rsk and longterm debt as wth bankruptcy costs a larger varance n earnngs mples lower leverage although, as noted by Cassar and Holmes (2003), earler emprcal results obtaned usng SME samples have been somewhat mxed. We a pror expect that the negatve relatonshp holds true n partcular n the case of long-term debt, but not necessarly n the case of short- 6 Based on the F-test the equaltes of the coeffcents for PP&E and INV were rejected. These results are not reported but avalable upon request.

10 9 term debt snce the hgher rsk especally may reduce the long-term lendng capacty of a small frm. Proftablty and captal structure can be hypotheszed to have a negatve relatonshp snce proftable frms wth access to retaned profts can use these for frm fnancng nstead of outsde sources (see, e.g., Wald 1999, Cassar and Holmes 2003, and Hall et al. 2004). In other words, there s a peckng order for the funds wth hgher preferences to use those generated wthn the busness. In lne wth earler studes such as those by Wald (1999) and Cassar and Holmes (2003), return on assets (ROA) s used as the performance measure. We also nclude a fve-year average of sales growth (GROWTH) to control for the effects of realzed frm growth on debt to assets ratos. Lke Wald (1999), Cassar and Holmes (2003), and Hall et al. (2004) frm sze and lke Hall et al. (2000, 2004) frm age measures are ncorporated nto the analyss. We also defne the frm sze varable (ln(ta)) as the natural logarthm of the book value of total assets and the frm age varable as the natural logarthm of the number of years snce the frm s foundaton (ln(age)). Accordng to Harrs and Ravd (1991), leverage ratos tend to vary across ndustres. Smlar emprcal evdence usng SME data s also provded n Hall et al. (2000) examnng the ndustry effects on captal structures. Consstent wth the suggestons of these studes, we nclude ndustry dummes to control for varatons across ndustres Sample statstcs Table 1 reports the descrptve statstcs and prelmnary data analyss. Panel A reports the means, medans, standard devatons, and mnmum and maxmum values for the varables used n the analyss. As can be seen from the table, the mean for the rato of long-

11 10 term debt to the book value of total assets (LTD) s 15.6 % and the mean for the rato of short-term debt to total assets (STD) s 41.6 %. The correspondng medan values appear to be somewhat lower, 7.1 % and 37.6 %, respectvely. It may be observed that these leverage ratos are n general well n lne, for example, wth Cassar and Holmes (2003) and Hall et al. (2004) whch use the Australan and European SME samples, respectvely. Panel B separates regular frms (43.3 % of all frms) from the frms wth growth strateges (56.7 % of all frms). Panel B presents the means and medans separately for both groups and t- and the Wlcoxon sgned rank test statstcs to test dfferences n means and medans, respectvely. Several nterestng observatons emerge from ths panel. Comparson of growth orented and regular frms suggests that growth orented frms are more leveraged than regular frms n terms of both long-term and short-term debt. Growth orented frms have also less property, plant and equpment, but are larger and younger than regular frms. Furthermore, t can be observed that frms wth growth strateges have actually experenced sgnfcantly hgher realzed growth. Interestngly, return on assets s not found to be statstcally dfferent between growth orented and regular frms. Panel C presents the Pearson correlaton coeffcents for the varables. As can be seen from Panel C, both short and long-term debt ratos are postvely correlated wth growth strateges and negatvely correlated wth ROA. The long-term debt rato appears to be postvely correlated wth varables PP&E, DEPR, and ln(age) and negatvely correlated wth varables RISK, INV, and GROWTH. For the short-term debt rato, the correlaton coeffcents wth correspondng varables have opposte sgns, and, as can be seen, the short and long-term debt ratos are negatvely correlated. In sum, the reported statstcs are very smlar, for example, to those of Cassar and Holmes (2003).

12 11 (Insert Table 1 about here.) 3. Impact of growth strateges on captal structure 3.1. Effects of growth strateges on the long and short-term debt ratos Both the standard OLS estmator (e.g. Hall et al. 2000, Cassar and Holmes 2003, and Hall et al. 2004) and the hetoroskedastc tobt estmator (Wald 1999) have prevously been used to nvestgate the nfluences on captal structures. The basc structure of our multvarate analyss s smlar, for example, to Wald (1999), Hall et al. (2000, 2004), and Cassar and Holmes (2003). Thus, we begn our analyss by frst estmatng the followng OLS regresson to examne the mpact of growth strateges on the long-term debt rato LTD : LTD 5 k 0 + α k SIC β 0 k= 1 = α + Growth _ Strategy + δ RISK + PP & (1) 1 δ 2 E INV δ 3 DEPR ROA 6 GROWTH + + δ 4 + δ δ + δ ln( TA ) + 8 ln( AGE ) + e, 7 δ where k SIC s a dummy varable for a frm accordng to the ndustry classfcaton, Growth _ Strategy s a dummy varable ndcatng whether the frm has a growth strategy, RISK s the standard devaton of the frst dfferences n return on assets, property, plant and equpment as a fracton of total assets, total assets, DEPR s the rato of deprecaton of total assets, PP & E denotes INV s the rato of nventores to ROA s return on assets,

13 12 GROWTH s a fve-year average of sales growth, ln( TA ) s the natural logarthm of the book value of total assets, and ln( AGE ) s the natural logarthm of the number of years snce the frm s foundng. It can be noted that the basc set of control varables s the same as n Wald (1999). In the model estmaton, the possblty of multcollnearty s detected usng the varance nflaton factors and t s concluded that multcollnearty s not a problem n the current analyss (see Judge, Hll, Grffths, Lütkepohl and Lee 1988, pp ). Based n the Whte test, t s concluded that the error varances are heteroskedastc. Consequently, the reported t-values are based on the Whte (1980) heteroskedastcty consstent covarance matrx. To take nto account the fact that the range of dependent varables s lmted, we ntroduce a partly latent varable y and defne that the long-term debt to totals assets rato LTD =, f y > 0 and LTD = 0 f y 0. Followng Wald (1999), we then estmate y the followng censored regresson model: y 5 k 0 + α k SIC β 0 k = 1 = α + Growth _ Strategy + δ RISK + PP & (2) 1 δ 2 E + δ 3 INV + δ 4 DEPR + δ 5 ROA5 + δ 6 GROWTH + δ ln( TA ) + 8 ln( Frm _ Age ) + e, 7 δ where the explanatory varables are defned as n the case of model (1). The motvaton for the use of the tobt estmator arses from the fact that the dependent varables, debt to total asset ratos are censored at zero. Snce the lkelhood rato test soundly rejects the null hypothess of homoskedastc error varances n all cases, we employ a multplcatve error specfcaton as n Wald (1999) (see e.g. Judge et al. 1988, p. 365).

14 13 Table 2 reports the results of the OLS and tobt regressons n whch long-term debt ratos of SMEs are explaned by the growth strategy varables and the outlned set of control varables. As can be seen from Panel A of Table 2, after controllng for ndustry effects and frm characterstcs the long-term debt rato s sgnfcantly postvely related to growth strategy, ndcatng that growth orented frms tend to be more leveraged n terms of longterm debt. Our multvarate results reported n Table 2 are therefore consstent wth the unvarate results n Panel B of Table 1. Panel A of Table 2 moreover suggests that growth strateges: (1) new product development, (2) ncreasng export, (3) networkng, and (4) marketng and sales promoton are postvely sgnfcantly related to the long-term debt rato. In partcular, new product development, ncreasng export and networkng appear to requre more long-term debt. The results regardng the control varables are consstent wth earler studes (see, e.g., Wald 1999, Hall et al. 2000, Cassar and Holmes 2003, and Hall et al. 2004). Of the control varables, PP&E and INV are postvely sgnfcantly related to long-term debt, whereas the control varables ROI and ln(ta) are negatvely sgnfcantly related. It may be observed that the OLS and tobt regressons produce very smlar estmates. Accordng to the earler lterature leverage ratos of tend vary across ndustres (see, e.g., Harrs and Ravd 1991, and Hall et al. 2000). Panel B of Table 2 reports the results of examnng the effects of dfferent ndustres. As can be seen from the panel, ndustres (F) constructon, (G) wholesale and retal trade; repar of motor vehcles, motorcycles and personal and household goods, and (K) real estate, rentng and busness actvtes have less long-term debt than the remanng ndustres. Our fndngs seem to be well n lne wth Hall et al. (2000). Fnally, Panel C of Table 2 reports the results of the model specfcaton tests and heteroskedastcty estmates. As can be observed, the adjusted R 2 s 0.35 for both OLS

15 14 regressons. F-tests show that the regressons are statstcally hghly sgnfcant. For the tobt regressons, accordng to the lkelhood rato test, homoskedastcty of error terms s soundly rejected and consequently, followng Wald (1999), an approprate multplcatve specfcaton s chosen. The lkelhood rato test clearly ndcates that the selected heteroskedastcty specfcaton s more approprate than the homoskedastcty assumpton. (Insert Table 2 about here.) As the next step, we examne the mpact of growth strateges on the short-term debt rato usng smlar OLS and tobt regressons. Table 2 reports the results of these regressons n whch the short-term debt rato STD s explaned by the strategy varables and the set of control varables. Panel A of Table 3 shows that the short-term debt rato, lke the long-term debt rato, s sgnfcantly postvely related to growth strategy thereby suggestng that growth orented frms tend to be more leveraged. Panel A also shows that the relatonshp between the short-term debt rato and the growth strateges, (3) networkng, and (4) marketng and sales promoton are postve and sgnfcant at the 5 % level. As can also be seen, the control varables, INV, DEPR, and GROWTH appear to be postvely related to the amount of short-term debt. In contrast, the varables PP&E, ROA, ln(ta) and ln(age) are negatvely sgnfcantly related to the short-term debt rato. In the case of short-term debt rato, the results from the OLS and tobt regressons are also smlar, ndcatng that estmaton results are robust. Panel B of Table 3 presents varaton of short-term debt across ndustres. Of the ndustres, frms n (F) constructon, and (I) transport, storage and communcaton ndustres, n partcular, have more short-term debt than the frms n other ndustres. Panel C of Table 3

16 15 presents the results of the model specfcaton tests and heteroskedastcty estmates. The adjusted R 2 s 0.16 for all the OLS regressons, and on the bass of the F-test t s concluded that the varables n the regressons are jontly statstcally hghly sgnfcant. Accordng to the lkelhood rato test, homoskedastcty of error terms s rejected and consequently, a consstent multplcatve specfcaton s ftted. (Insert Table 3 about here.) 3.2. Further analyss of growth strateges We contnue our analyss of the growth strateges by estmatng user profles frst jontly for the growth orented frms and then separately for each growth strategy. The estmatons are based on the followng probablty model: Pr( Strategy) = Φ( z ), (3) 5 k z = α 0 + α k SIC + β1 LTD + β 2 k = 1 STD + δ RISK + PP & 1 δ 2 E + δ 3 INV + δ 4 DEPR + δ 5 ROA5 + δ 6 GROWTH + δ ln( TA ) + δ 8 ln( Frm _ Age ), 7 where k SIC s a dummy varable for a frm accordng to the ndustry classfcaton, s the rato of long-term debt to total assets, assets, LTD STD s the rato of short-term debt to total RISK the standard devaton of the frst dfferences n return on assets, PP & E

17 16 denotes property, plant and equpment as a fracton of total assets, INV s the rato of nventores to total assets, DEPR s the rato of deprecaton of total assets, ROA s return on assets, GROWTH s a fve-year average of sales growth, ln( TA ) s the natural logarthm of the book value of total assets, and ln( Frm _ Age ) s the natural logarthm of the number of years snce the frm s foundng. Table 4 reports the results of the probt models. As can be seen, after controllng for ndustry effects and frm characterstcs, both the rato of long-term debt to total assets LTD ) and the rato of the short-term debt to total assets ( STD ) appear to be dscrmnatng ( factors between the growth orented and regular frms at the 5% sgnfcance level. The results also ndcate that growth orented frms have less PP & E but more deprecatons. Furthermore, these multvarate results confrm that the growth orented frms are both larger and younger than the regular frms. It may also be noted that the growth orented frms have grown n the past. Separatng dfferent strateges shows that the rato of long-term debt to total assets ( LTD ) seems to be a dscrmnatng factor n cases of ncreasng export, networkng, and marketng and sales promoton strateges. The short-term debt to total assets ( STD ) s a dscrmnatng factor for the networkng, and marketng and sales promoton strateges. PP & E appears to be negatvely related to the probablty of a frm followng exportng, networkng and sales promoton and marketng strateges. In contrast, realzed sales growth s postvely related to the lkelhood of the new product development strategy. As can be seen from the table, frm age ncreases the use of growth strateges. Furthermore, networkng, and marketng and sales promoton strateges seem to be negatvely related to frm age.

18 17 (Insert Table 4 about here.) 4. Conclusons We examne a relatonshp between growth strateges and captal structures usng the fnancal statements of 1,153 randomly selected Fnnsh non-lsted ncorporated small and medum-szed enterprses (SMEs). We dentfy the frms wth the followng four growth strateges: (1) new product development, (2) ncreasng export, (3) networkng, and (4) marketng and sales promoton. Our emprcal results suggest that of these strateges, the networkng, and marketng and sales promoton strateges are typcal for younger frms, whereas ncreasng export s characterstc for older frms. Furthermore, the use of growth strateges tends ncrease wth frm age. As our prmary results, both parametrc and nonparametrc unvarate comparsons and a multvarate statstcal analyss controllng for varous frm characterstcs such as ndustry varaton, asset structure, frm age and sze consstently show that growth orented frms wth growth strateges have more both long-term and short-term debt than comparable frms wth no growth strateges. We also fnd that the type of growth strategy s related to the amount of long-term and short-term debt. Networkng and, n partcular, marketng and sales promoton typcally ncrease the use of short-term debt, whereas new product development and export actvtes seem to requre more long-term fnancng. Fnally, consstent wth earler studes, we fnd that a frm s captal structure depends on ts ndustry, proftablty, rsk, growth, asset structure, sze and age.

19 18 References Ang, J. S. (1991). Small busness unqueness and the theory of fnancal management. Journal of Small Busness Fnance 1, Ang, J. S., R. A. Cole and J. W. Ln (2000). Agency costs and ownershp structure. Journal of Fnance 55, Berger, A.N. and G.F. Udell (1995). Relatonshp lendng and lnes of credt n small frm fnance. Journal of Busness 68, Brounen, D., A. de Jong, K. Koedjk (2006). Captal structure polces n Europe: Survey Evdence. Journal of Bankng and Fnance 30, Cassar, G. and S. Holmes (2003). Captal structure and fnancng of SMEs: Australan evdence. Accountng and Fnance 43, Cavalluzzo, K.S., L.C. Cavalluzzo and J.D. Wolken (2002). Competton, small busness fnancng, and dscrmnaton: Evdence form a new survey. Journal of Busness 75, Dens, D.J. (2004). Entrepreneural fnance: an overvew of the ssues and evdence. Journal of Corporate Fnance 10,

20 19 Esenberg, T., S. Sundgren and M. T. Wells (1998). Larger board sze and decreasng frm value n small frms. Journal of Fnancal Economcs 48, Hall, G., P. Hutchnson and N. Mchaelas (2000). Industry effects on the determnants of unquated SMEs captal structure. Internatonal Journal of Economcs and Busness 7, Hall, G., P. Hutchnson and N. Mchaelas (2004). Determnants of the captal structures of European SMEs. Journal of Busness Fnance & Accountng 31, Harrs, M. and A. Ravd (1991). The theory of captal structure. Journal of Fnance 46, Judge, G. G., T-C. Lee and R. C. Hll (1985). Introducton to the theory and practce of econometrcs. John Wley & Sons. Korajczyk, R.A. and A. Levy (2003). Captal structure choce: macroeconomc condtons and fnancal constrants. Journal of Fnancal Economcs 68, McFadden, D. (1974). Condtonal logt analyss of qualtatve choce behavor. In Fronters n Econometrcs, ed. P. Zarembka, New York: Academc Press. McKelvey, R.D. and W. Zavona (1975). A statstcal model for the analyss of ordnal level dependent varables. Journal of Mathematcal Socology 4,

21 20 Myers, S.C. (1984). The captal structure puzzle. Journal of Fnance 34, Myers, S.C. and N.S. Majluf (1984). Corporate fnancng and nvestment decsons when frms have nformaton that nvestors do not have. Journal of Fnancal Economcs 13, Rajan, R.G. and L Zngales (1995). What do we know about captal structure? Some evdence from nternatonal data. Journal of Fnance 50, Wald, J.K. (1999). How frm characterstcs affect captal structure: an nternatonal comparson. Journal of Fnancal Research 22, Whte, H. (1980). A heteroscedastcty-consstent covarance matrx estmator and a drect test for heteroscedastcty. Econometrca 48,

22 21 Table 1. Descrptve data for growth and non-growth frms (n=1,153 frms) Ths table reports the descrptve statstcs and prelmnary data analyss. Panel A gves the means, medans, standard devatons, and mnmum and maxmum values for the followng varables: LTD s the long-term debt to total assets rato, STD s the short-term debt to total assets rato, RISK s the standard devaton of the frst dfferences n return on assets, PP&E denotes property, plant and equpment as a fracton of total assets, INV s the rato of nventores to total assets, DEBR s the rato of deprecaton of total assets, ROA s return on assets, GROWTH s the fve-year average of sales growth, ln(ta) s the natural logarthm of the book value of total assets, and ln(age) s the natural logarthm of the number of years snce the frm s foundng. Panel B separates regular frms (43.3 % of all frms) from the growth orented frms wth growth strateges (56.7 % of all frms). It presents the means and medans separately for both groups and t- and the Wlcoxon sgned rank test statstcs, respectvely. Panel C reports the Pearson correlaton coeffcents for the key varables. Panel A. Summary statstcs for all frms Varable Mean Medan Std.dev. Mn Max Total Assets ( 1,000) Number of Employees Frm Age (n years) LTD STD RISK PP&E INV DEPR ROA GROWTH Panel B. Dfferences of means and medans tests Means Medans Regular Growth t-value Regular Growth Z-value LTD * STD *** *** RISK PP&E *** *** INV *** * DEPR ROA GROWTH ** *** Ln(TA) *** *** Ln(AGE) * **

23 22 Panel C. Pearson correlaton coeffcents for selected varables Varable LTD STD Strategy RISK PP&E INV DEPR ROA GROWTH lnta STD Strategy RISK PP&E INV DEPR ROA GROWTH Ln(TA) Ln(AGE)

24 23 Table 2. Effects of growth strateges on long-term debt ratos Panel A reports the estmaton results of the OLS and heteroskedastc tobt regressons n whch the long-term debt to total assets rato s explaned by a growth strategy dummy varable (Strategy) or the followng growth strateges: (1) new product development (Products), (2) ncreasng export (Export), (3) Networkng, and (4) marketng and sales promoton (Marketng). Control varables are as follows: RISK s the standard devaton of the frst dfferences n return on assets, PP&E denotes property, plant and equpment as a fracton of total assets, INV s the rato of nventores to total assets, DEBR s the rato of deprecaton of total assets, ROA s return on assets, GROWTH s a fve-year average of sales growth, ln(ta) s the natural logarthm of the book value of total assets, and ln(age) s the natural logarthm of the number of years snce the frm s foundng. Panel B reports the ndustry effects. The ndustres are (D) manufacturng, (F) constructon, (G) wholesale and retal trade; repar of motor vehcles, motorcycles and personal and household goods, (H) hotels and restaurants, (I) transport, storage and communcaton, and (K) real estate, rentng and busness actvtes. Panel C presents the model specfcaton tests and heteroskedastcty estmates. t- values n the OLS regressons are based on the Whte (1980) heteroskedastcty consstent covarance matrx. Panel A. Determnants of captal structure OLS Tobt OLS Tobt Strategy *** *** (3.25) (2.65) Products * ** (1.73) (2.07) Export *** *** (2.60) (3.03) Networkng *** * (2.72) (1.87) Marketng ** (1.97) (1.37) RISK * * (-0.37) (-1.68) (-0.51) (-1.85) PP&E *** *** *** *** (17.26) (17.40) (17.30) (17.43) INV *** *** *** *** (7.43) (9.07) (7.48) (9.04) DEPR (-1.07) (0.42) (-1.04) (0.42) ROA *** *** *** *** (-4.35) (-6.55) (-4.47) (-6.58) GROWTH (0.60) (1.14) (0.66) (1.17) Ln(TA) *** *** *** *** (-3.89) (-3.87) (-4.08) (-4.10) Ln(AGE) (-1.20) (-0.13) (-1.26) (-0.30)

25 24 Panel B. Industry effects OLS Tobt OLS Tobt Intercept *** *** * (3.85) (1.64) (4.01) (1.86) F *** *** *** *** (-4.46) (-3.30) (-4.22) (-2.93) G *** ** ** * (-2.59) (-2.34) (-2.11) (-1.81) H (-0.37) (-0.44) (-0.20) (-0.20) I ** * (-1.27) (-1.96) (-1.19) (-1.72) K *** *** *** *** (-4.53) (-4.54) (-4.28) (-4.25) Panel C. Model specfcaton tests and heteroskedastcty estmates OLS Tobt OLS Tobt F-value *** *** Adjusted R LR test *** *** HET *** *** (3.38) (3.35) HET *** *** (3.51) (3.45) HET *** *** (3.32) (3.21) HET *** *** (-3.12) (-3.17)

26 25 Table 3. Effects of growth strateges on short-term debt ratos Panel A reports the estmaton results of the OLS and heteroskedastc tobt regressons n whch the short-term debt to total assets rato s explaned by a growth strategy dummy varable (Strategy) or the followng growth strateges: (1) new product development (Products), (2) ncreasng export (Export), (3) Networkng, and (4) marketng and sales promoton (Marketng). Control varables are as follows: RISK s the standard devaton of the frst dfferences n return on assets, PP&E denotes property, plant and equpment as a fracton of total assets, INV s the rato of nventores to total assets, DEBR s the rato of deprecaton of total assets, ROA s return on assets, GROWTH s the fve-year average of sales growth, ln(ta) s the natural logarthm of the book value of total assets, and ln(age) s the natural logarthm of the number of years snce the frm s foundng. Panel B reports the ndustry effects. The ndustres are (D) manufacturng, (F) constructon, (G) wholesale and retal trade; repar of motor vehcles, motorcycles and personal and household goods, (H) hotels and restaurants, (I) transport, storage and communcaton, and (K) real estate, rentng and busness actvtes. Panel C presents the model specfcaton tests and heteroskedastcty estmates. t- values n the OLS regressons are based on the Whte (1980) heteroskedastcty consstent covarance matrx. Panel A. Determnants of captal structure OLS Tobt OLS Tobt Strategy ** ** (2.17) (2.36) Products (0.20) (0.27) Export (-1.25) (-0.35) Networkng ** * (2.13) (1.80) Marketng ** *** (2.45) (2.72) RISK (0.43) (1.49) (0.54) (1.62) PP&E *** *** *** *** (-8.31) (-9.50) (-8.27) (-9.43) INV *** *** *** *** (4.13) (3.83) (4.26) (3.92) DEPR *** *** *** *** (2.96) (3.37) (2.94) (3.37) ROA *** *** *** *** (-3.65) (-6.35) (-3.68) (-6.36) GROWTH * * (1.67) (1.39) (1.74) (1.38) Ln(TA) ** *** * ** (-2.08) (-2.64) (-1.82) (-2.42) Ln(AGE) *** *** *** *** (-2.87) (-3.24) (-2.70) (-3.11)

27 26 Panel B. Industry effects OLS Tobt OLS Tobt Intercept *** *** *** *** (8.30) (9.07) (8.08) (8.81) F ** ** * (2.02) (2.16) (1.62) (1.91) G (-0.07) (-0.08) (-0.46) (-0.37) H (0.55) (0.02) (0.34) (-0.11) I *** *** *** *** (4.00) (4.20) (3.61) (3.96) K (0.56) (0.45) (0.27) (0.30) Panel C. Model specfcaton tests and heteroskedastcty estmates OLS Tobt OLS Tobt F-value *** *** Adjusted R LR test *** *** HET *** *** (4.04) (4.02) HET *** *** (5.45) (5.34) HET *** *** (-5.79) (-5.72) HET (-1.11) (1.01)

28 27 Table 4. Debt to total assets ratos as dscrmnatng factors Table 4 reports the estmaton results of a probt model n whch the followng growth strateges: (1) new product development (Products), (2) ncreasng export (Export), (3) Networkng, and (4) marketng and sales promoton (Marketng). Explanatory varables are as follows: LTD s the long-term debt to total assets rato, STD s the short-term debt to total assets rato, RISK s the standard devaton of the frst dfferences n return on assets, PP&E denotes property, plant and equpment as a fracton of total assets, INV s the rato of nventores to total assets, DEBR s the rato of deprecaton of total assets, ROA s return on assets, GROWTH s the fve-year average of sales growth, ln(ta) s the natural logarthm of the book value of total assets, and ln(age) s the natural logarthm of the number of years snce the frm s foundng. The ndustres are (D) manufacturng, (F) constructon, (G) wholesale and retal trade; repar of motor vehcles, motorcycles and personal and household goods, (H) hotels and restaurants, (I) transport, storage and communcaton, and (K) real estate, rentng and busness actvtes. Frequency s the proporton of growth orented frms. Pseudo R 2 s the McKelvey and Zavona (1975) R Square. LRI s the McFadden (1974) lkelhood rato ndex. Growth (1) (2) (3) (4) Varable Strategy Product Export Network Market Intercept *** *** *** *** *** (-3.56) (-3.64) (-5.33) (-3.13) (-2.69) F *** *** *** ** *** (-5.49) (-4.05) (-4.21) (-2.14) (-3.67) G *** * *** (0.77) (0.44) (-3.25) (-1.66) (2.73) H (-0.81) (0.06) (0.00) (-1.50) (0.14) I * * (-0.91) (-1.72) (-1.63) (1.69) (-1.06) K * (1.07) (0.75) (-1.85) (0.85) (1.61) LTD *** ** *** * (3.28) (1.50) (2.54) (2.84) (1.85) STD ** ** *** (2.32) (0.34) (-1.08) (2.32) (2.68)

29 28 Table 4. contnued Growth (1) (2) (3) (4) Varable Strategy Product Export Network Market RISK (-0.81) (-0.16) (1.03) (-0.05) (-1.47) PP&E *** * *** *** (-3.68) (-1.18) (-1.77) (-2.71) (-2.78) INV (0.26) (0.89) (0.95) (-1.39) (0.47) DEPR ** ** (2.01) (0.86) (0.77) (0.45) (1.98) ROA (-0.25) (-1.10) (-0.48) (1.18) (-0.75) GROWTH ** * ** (2.26) (1.90) (1.16) (0.67) (2.27) Ln(TA) *** *** *** *** *** (5.29) (3.04) (5.15) (3.37) (3.21) Ln(AGE) * ** * (-1.87) (-0.16) (0.07) (-2.19) (-1.82) Obs. 1, Frequency Pseudo R LRI