The Evolution of Financing Structure in U.S. Startups

Transcription

1 The Journal of Entrepreneural Fnance Volume 19 Issue 1 Sprng 2017 Artcle 4 February 2017 The Evoluton of Fnancng Structure n U.S. Startups Carmen Cote Unversty of Hartford Joseph Farhat Central Connectcut State Unversty Follow ths and addtonal works at: Part of the Entrepreneural and Small Busness Operatons Commons, and the Fnance and Fnancal Management Commons Recommended Ctaton Cote, Carmen and Farhat, Joseph (2017) "The Evoluton of Fnancng Structure n U.S. Startups," The Journal of Entrepreneural Fnance: Vol. 19: Iss. 1, pp. -. Avalable at: Ths Artcle s brought to you for free and open access by the Grazado School of Busness and Management at Pepperdne Dgtal Commons. It has been accepted for ncluson n The Journal of Entrepreneural Fnance by an authorzed edtor of Pepperdne Dgtal Commons. For more nformaton, please contact josas.bartram@pepperdne.edu, anna.speth@pepperdne.edu.

2 The Evoluton of Fnancng Structure n U.S. Startups Carmen Cote 1 Unversty of Hartford Barney School of Busness Joseph Farhat Central Connectcut State Unversty School of Busness Abstract In ths artcle we examne how startup busnesses fnance ther operatons over tme. We employ the Latent growth modelng technque to test the fnancal growth cycle theory developed by Berger and Udell (1998). The data used n ths study s the Kauffman Frm Survey, the largest longtudnal data set comprsed of a random sample of U.S. startups launched n 2004 and surveyed annually through Consstent wth the predctons of fnancal growth cycle theory, n the startup stage, entrepreneurs rely on ntal nsder captal sources such as personal savngs, fnancng offered by frends and famly, quas-equty, and personal debt. Over tme, as busnesses become less opaque, the proporton of busness debt and trade credt fnancng n total captal njecton volume ncreases sgnfcantly. Busnesses wth hgh R&D actvty and those that possess ntellectual property rghts fnance ther operatons predomnantly wth equty - partcularly external equty rased from angels and venture captalsts, and busness debt - partcularly bank loans and credt lnes. Owner s educaton and race have a sgnfcant mpact on the type of captal njectons over the busness lfe cycle. Hghly educated owners choose to nject lower proportons of personal debt and trade fnancng, whereas whte owners nject lower proportons of personal equty and rely more on trade fnancng. Keywords: startups, fnancng structure; fnancal growth cycle theory; R&D ntensty; ntellectual property rghts; survey methodology. JEL classfcatons: C83, G32, M13, O31, O32 1 Carmen Cote gratefully acknowledges summer research support from the Barney School of Busness. 1

3 1. Introducton Pror research has explored the captal structure decsons of large, well-establshed publcly-traded frms. To date, however, very few extant studes have examned the dynamcs of captal structure of newly formed busnesses. The dynamcs of captal structure refer to ways n whch a busness fnances ts overall operatons and growth over tme, as needs for dfferent sources of funds such as debt, equty or a combnaton of the two dffer n progressve stages of ts development. In ths artcle, we seek to explore how newly formed busnesses fnance ther operatons from ncepton to the later stages of ther development. To do so, we use the fnancal growth cycle theory developed by Berger and Udell (1998) and emprcally examne whether the captal structures of newly formed frms follow the predctons of ths theory. Smultaneously, we also examne whether the peckng-order and trade-off theores explan the changes n captal structure of startup frms from ncepton to later stages of development. Our data s sourced from the Kauffman Frm Survey 2 (KFS), the largest longtudnal data set comprsed of 4,928 U.S. startups launched n 2004 and surveyed annually through The KFS provdes a wealth of detaled nformaton on both new frms and ther owners, ncludng performance data and nformaton on fnancng sources and strateges as well as personal characterstcs of the entrepreneur (Ballou et al., 2008). Accordng to the evdence yelded by pror research n ths feld, n ther attempts to rase debt and equty, new and small frms face dfferent challenges than publcly held frms do. Extant research has shown that startup captal, as well as future captal njectons, are crucal to frm survval. Most studes on captal structure that employed the KFS data have focused on the determnants of captal structure or the fnancng choces n the startup year. Thus, ther authors dd not use the full panel data set ( ) as t was not avalable at the tme the work was conducted (Coleman and Robb, 2012; Robb and Robnson, 2014; Coleman, Cote, and Farhat, 2016). Whle offerng valuable fndngs, these studes largely gnore the sources of fundng employed by a frm at varous stages of ts lfe cycle. Gven that the full panel data and the mputed data of the KFS are now avalable, t s tmely to examne the trajectores of small busnesses fnancal structures and examne the determnants of varous sources of captal njectons n newly formed frms. The approach we adopt n ths work s evolutonary, as t ams to ascertan how newly formed frms captal structures evolve over several years followng ther ncepton. Our results pont to a sgnfcant fnancal growth pattern n the use of varous sources of debt and equty durng the lfe cycle of the busness. Consstent wth Berger and Udell s model, we fnd that, at the startup stage, entrepreneurs rely on ntal nsder captal sources such as personal savngs, fnancng provded by frends and famly, quas-equty, and personal debt. Over tme, as busnesses become less opaque, the proporton of busness debt and trade credt fnancng n the total captal njecton volume ncreases sgnfcantly. We further note that, although the proporton of owner s equty n total captal njectons decreases over tme, the annual balance of owner s equty ncreases, suggestng that owners use retaned earnngs to ncrease ther ownershp stake n the frm. Several busness and owner characterstcs explan the type and magntude of varous types of captal njectons. Busnesses wth hgh nformaton asymmetry rely more on equty njectons, especally owner s personal savngs, whereas those undertakng nnovatve actvtes are more 2 Certan data ncluded heren are derved from the Kauffman Frm Survey restrcted-access data fle. Any opnons, fndngs, and conclusons or recommendatons expressed n ths materal are those of the author(s) and do not necessarly reflect the vews of the Ewng Maron Kauffman Foundaton. 2

4 lkely to fnance ther operatons wth outsde equty njectons provded by angels and venture captalsts. A hgher proporton of tangble assets explans the greater use of debt njectons, both personal and busness debt sources, whereas hgher level of sales drves busnesses to use both debt njectons and trade fnance and rely less on equty njectons. Startups owned by two or more owners tend to nject more equty, especally outsde equty and less debt nto the captal structure of ther busness. Educaton and race also have a sgnfcant mpact on the captal njectons over the busness lfe cycle. Hghly educated owners choose to nject lower proportons of personal debt and trade fnancng, whereas whte owners nject lower proportons of personal equty and rely more on trade fnancng. In provdng the aforementoned fndngs, we contrbute to the current stream of lterature n three mportant ways. Frst, our results provde nsghts nto the ways n whch fnancal theory explans the captal njectons n startup busnesses. Usng a Latent growth model to examne the systematc change n the fnancal structure of newly formed busnesses we show that there s a sgnfcant fnancal growth pattern n the use of varous sources of debt, equty and trade fnance durng the lfe cycle of the frm. Second, ths artcle bulds upon and contrbutes to the exstng body of work carred out on small busness fnancng n the startup year. Fnally, our research makes use of a comprehensve data set - the Kauffman Frm Survey whch represents a random sample of U.S. busnesses launched n 2004, proportonally weghted across specfc subpopulatons. Therefore, the bas that often emerges from small or under-representatve samples s completely removed. The remander of ths artcle s organzed as follows. Ths bref ntroducton s followed by a secton that presents our theoretcal framework and stuates our work n the context of pertnent research. Our data, ncludng the characterstcs of the Kauffman Frm Survey, methodology and results, are descrbed n Secton 3. Fnally, Secton 4 provdes a summary of our fndngs and concludes the paper. 2. Theoretcal Framework and Hypotheses 2.1. Lfe Cycle Theory of the Frm The fnancal growth lfe cycle model developed by Berger and Udell (1998) s based on the dea that frms have dfferent fnancal needs and optons as they grow and become less nformatonally opaque. Therefore, when facng a sze/age/nformaton contnuum, smaller, younger, more opaque frms should predomnantly rely on nsder fnance, trade credt and/or angel fnance. On the other hand, as frms grow and become less opaque, they should rely more on ntermedated fnance, both debt (bank loans) and equty (venture captal). The fnancal growth lfe cycle model conceptualzes the fact that nformaton opacty over the lfe cycle of the frm creates fnancal peckng order, suggestng that, n early stages of the frm s lfe, the entrepreneur reles on ntal nsder fnancal sources (.e., personal savngs, loans from frends and famly, quas-equty, personal debt, and busness debt), trade credt, and angel fnance, whereas, at a later stage, frm gans access to external debt and equty and therefore, personal fundng becomes relatvely less mportant. The fnancal growth lfe cycle model predcts that the major captal structure varables wll follow a specfc trajectory over the lfe of the busness. In lne wth ths vew, we hypothesze the followng: H1(a): In the startup stage, owner s personal savngs and personal debt are the most mportant sources of fnancng. H1(b): Over tme, the relatve mportance of owners equty and personal debt n total fnancng wll declne, whle the relatve mportance of busness debt and trade credt n total 3

5 fnancng wll ncrease The Entrepreneur s Personal Characterstcs The lterature on ndvdual rsk-takng behavor reveals that demographc and socoeconomc factors, such as age, gender, educaton, experence, and wealth, have a sgnfcant nfluence on ndvdual rsk tolerance. McInsh (1982) and Bertrand and Schoar (2003) show that rsk takng decreases wth age and ncreases wth educaton, ncome, wealth, experence, and sophstcaton. In addton, the entrepreneur s gender may have an mpact on type and sze of ther startup fnancng sources as males tend to be more rsk tolerant than females. The emprcal evdence on the mpact of owners characterstcs on startup fnancng has mxed results. Usng a panel of 2,000 Dutch entrepreneurs, Verheul and Thurk (2001) show that, even though female entrepreneurs start wth lower levels of startup captal, the proporton of debt and equty n total captal s smlar to that of male entrepreneurs. For a sample of startups from 27 countres, Nofsnger and Wang (2011) report that experence n managng startups s the key factor n securng fundng, as experenced owners are able to reduce nformaton asymmetry and obtan fnancng n the frst year of operatons. Besdes experence, educaton s another trat found to ncrease the lkelhood of usng credt n the startup year (Cole and Sokolyk, 2013). However, for a sample of Australan startups, Cassar (2004) reveal that owner s demographc and socoeconomc characterstcs are rrelevant for captal structure decsons. The entrepreneur s personal wealth may also affect the composton of captal n small busnesses. For example, Cole et al. (2004) show that the commtment of a small busness owner s personal wealth s mportant to obtan external credt. Overall, these fndngs attest to the mpact of owner characterstcs, such as atttudes and belefs, educaton, and experence, on fnancal decson-makng. Thus, we hypothesze: H2: The entrepreneur s personal characterstcs nfluence the type of captal njectons n startup frms The Peckng Order Model The peckng order theory (Myers and Majluf, 1984; Myers, 1984) s based on the dea of nformaton asymmetry between the busness and ts potental nvestors. In other words, t posts that small busness owners/managers know more about the true value of the frm and the frm s rskness than do less nformed supplers of captal. Thus, to avod the undernvestment problem nherent n frms wth nformatonal opacty, small busness owners wll seek to fnance new projects usng personal funds (nternal funds, personal debt) rather than busness debt or external equty. The peckng order theory also explans why small frms tend to prefer debt to equty captal f external fnancng s requred. Thus, unlke large, publcly-traded frms, small busnesses should exhbt a postve relatonshp between nformatonal asymmetry and the propensty for personal debt fnancng. Therefore, we hypothesze: H3: Startups characterzed by hgher nformaton opacty wll rely more on nternal equty and personal debt and less on busness debt. In lght of the undernvestment rsk, Myers (1977, 1984) also suggested that frms wth hgh growth optons use less debt n order to preserve ther debt capacty. Thus, we expect frms wth hgher growth optons to rely less on debt fnancng. Hence, we hypothesze: H4: Hgh growth optons startups those wth extensve research and development (R&D) actvty or those that possess ntellectual property rghts rely more on nternal and external equty and less on debt fnancng. 4

6 Frms possessng greater tangble assets are expected to have a hgher debt capacty and lower costs of fnancal dstress, because they have the opton of pledgng those assets as collateral. Consequently, frms wth greater tangble assets have the ablty to secure more debt and therefore, rely more on external busness debt. Smlarly, larger frms are often more dversfed and more proftable relatve to smaller frms. These characterstcs make them more attractve to lenders. In lght of that, f external fundng s needed, we expect a postve relatonshp between frm sze and the use of debt fnancng The Trade-off Model The tax beneft bankruptcy cost trade-off models (Baxter, 1967; DeAngelo and Masuls, 1980; Kraus and Ltzenberger, 1973; Robchek and Myers, 1966; Scott, 1976) predct that frms wll seek to mantan an optmal captal structure by balancng the benefts and the costs of debt. The benefts nclude the tax sheld, whereas the costs comprse of expected fnancal dstress costs, among others. Accordng to ths model, a frm s proftablty wll be negatvely related to the expected fnancal dstress costs, snce the probablty that frms wth hgher and more stable profts wll enter bankruptcy would be lower. In addton, larger frms typcally experence lower proft volatlty, and thus possess hgher debt ratos relatve to smaller frms. The exstng agency cost trade-off models (Jensen and Mecklng, 1976; Easterbrook, 1984; Stulz, 1990; Harrs and Ravv, 1990; Hart and Moore, 1995) consder the possble conflcts of nterests between the partes nvolved n the frm, such as managers and outsde nvestors. Overall, the trade-off and agency theores predct that frms mantan an optmum captal structure, where the margnal beneft of debt equals the margnal cost. In summary, for publcly traded frms, the trade-off models predct that hgher debt wll be assocated wth hgher proftablty, lower nondebt tax shelds, low growth opportuntes, hgh asset tangblty, and lower expected bankruptcy costs. Startup busnesses are less lkely to take advantage of tax nterest deductblty. The tax benefts are usually neglgble, as some busnesses generate lttle or no revenue n the frst year of operaton. In fact, pror research reveals that new, small frms are partcularly susceptble to the problems of fnancal dstress and falure. Hgh falure rates, coupled wth hgh montorng costs for the supplers of captal, make t dffcult for startups to rase debt. Therefore, the trade-off theory predcts that startup frms use less debt due to hgh probablty of falure. 3. Data and Methodology 3.1. Data The Kauffman Frm Survey (KFS) s the largest longtudnal survey of startup busnesses n the Unted States. The survey data contan detaled nformaton on 4,928 busnesses founded n 2004 and provde annual trackng nformaton of ther busness status and performance untl A busness started n 2004 was defned as a new, ndependent busness that was created by a sngle person or a team of people, the purchase of an exstng busness, or the purchase of a franchse. For each establshment, the database provdes nformaton about busness characterstcs, strategy and nnovaton, busness organzaton and HR benefts, busness fnances, work behavor, ownershp and demographcs of up to ten owners. Detaled nformaton on the busness ncludes sources and types of captal n the startup year and over tme, ntellectual property rghts (patents, trademarks, copyrghts), R&D actvty, physcal locaton, whether the busness provdes a product, servce or both, revenues, asset structure, sales, employment, and ndustry. Owners' characterstcs (up to ten actve-owner-operators per busness) nclude: years of work experence, prevous startup 5

7 experence, tme commtment and educaton, net worth, age, gender, race and ethncty. The defntons of varables used n ths study are presented n the Appendx. In response to the Kauffman Foundaton s nterest n understandng the dynamcs of hghtechnology, medum-technology, and women-owned busnesses, the KFS sample s stratfed based on ndustral technology level (Hgh-Tech, Medum-Tech, and Non-Tech sectors) oversamplng busnesses n hgh- and medum-tech sectors. It s a common practce n polcymakng surveys to oversample a key populaton subgroup n response to the small sze of that subgroup or for a specal nterest n that subgroup. It s mportant to notce that oversampled hghtechnology and medum-technology busnesses n the KFS am at ncreasng the precson of standalone and comparatve analyses of these sub-groups. Usng a dsproportonate stratfed samplng requres a weghtng scheme to make the KFS sample a representatve sample of the target populaton. In the baselne survey and to account for oversamplng, the ntal weghts were defned as the nverse of the probablty of selecton, whch was calculated n each stratum. Next, the ntal weghts were adjusted to compensate for the busnesses that dd not respond or could not be located. A logstc propensty model was used to determne the probablty of locatng a busness as well as to determne the propensty to respond n the survey. Gven both probabltes and ther nverses, the weghts were adjusted to count for the jont condtonal probablty that a busness was selected for samplng, was located, and responded n the survey. Fnally, to make the data even more representatve of the populaton a post-stratfcaton was appled. Detals of the use of proper weghts for the KFS sample are avalable n Farhat (2012) and Farhat and Robb (2014). All analyses n ths paper are performed usng the weghts and stratfcaton. Although the mssng observatons n the KFS are not sgnfcant, the Kauffman foundaton has released a multply mputed data verson of the KFS data. Used by statstcal agences, multple mputaton s a very popular methodology to handle nonresponse n publc use surveys and to reduce measurement errors. The mputed values are deally ndependent draws from the predctve dstrbuton of the mssng values condtonal on the observed values. The KFS multply mputed data nclude the orgnal data along wth fve complete-data mputatons (m=5). Usng specal procedures, analyzng multply mputed data requres estmatng the statstcal model for each mputed dataset whch s fve n the case of KFS (Farhat and Robb, 2013). All analyses n ths paper are performed on the KFS multply mputed dataset Descrptve Statstcs In Table 1 Panel A we report the man characterstcs of busnesses comprsng our sample. On average, about 61 percent of frms have lmted lablty (S-corporatons and LLC), whle the remanng 39 percent are organzed as propretorshps and partnershps. In addton, about 83 percent of frms operate n an urban locaton and half of the startups operate from owners' homes. The data also reveal that about 87 percent of startups provde servces, 48 percent provde products, and 35 percent provde both servces and products. Our nformaton asymmetry proxy s very volatle durng the perod. On average, 14 per cent of startups have no avalable nformaton to calculate the commercal credt score and are therefore classfed as hgh nformaton asymmetry frms. Startups sze, measured by ln(assets), has a growth rate of 10 percent, whereas sales growth rate s about 53 percent for the perod. Credt rsk score (a measure of frms rskness) decreased from 3.41 to 2.9 across the sample perod. Data also show that less than 20 percent of frms report any form of ntellectual property rghts (IP) such as copyrghts and patents or R&D actvty. A hgh proporton of frms (63.35 percent) ndcate that 6

8 they have a comparatve advantage (somethng unque and dstnctve) relatve to ther compettors n the startup year (2004). However, the proporton of frms wth comparatve advantages decreases to per cent n the last year of survey (2011). On average, the number of employees ncreases from 1.06 ndvduals per frm n the frst year of operaton to about 2.92 n About 41 percent of frms total assets are tangble assets. On average, frms reported about $22,067 n annual proft durng ther frst eght years of operaton. In Table 1 Panel B we show the dstrbuton of our sample across ndustres. Wth respect to ndustry stratfcaton, our sample conssts of 1.88 percent hgh-tech frms, percent medum-tech frms, and percent non-tech frms. Usng the North Amercan Industry Classfcaton System (NAICS), the majorty of frms n our sample are n the servce and retal trade sectors (71 per cent), whle the lowest percentage of frms operate n the manufacturng sector (5.84 per cent). Table 1 Panel C shows the detaled owners characterstcs. The data reveal that the average age of the entrepreneurs was forty-seven years across the sample perod ( ). The entrepreneurs n our sample had an average of years of ndustry experence. Owners devoted about 39 hours per week to ther busness and, n about 47 percent of startups, the entrepreneurs are hghly educated (havng earned college degree or hgher). Wth respect to gender and race, percent of frms were owned by male entrepreneurs, about 5 percent have Hspanc owner, and percent of the sample frms are owned by whte entrepreneurs. The KFS survey provdes nformaton on varous types of debt and equty fnancng sources. For the purpose of our analyses, we classfy the fnancng varables nto nsder and outsder debt or equty, based on the fact that nsders provdng captal have more nformaton about the busness than outsders do. Insder s personal debt conssts of personal debt from famly and others, whereas personal credt cards and personal bank loans are consdered outsder s personal debt. In the same ven, nsder s busness debt comprses of busness debt from famly, employees, and other ndvduals. On the other hand, outsder s busness debt conssts of credt cards, bank loans, government s loans, and loans from other busnesses. Owner s equty conssts of equty provded by the owner(s). Insder s equty conssts of equty provded by spouse and/or parents, whereas outsder s equty ncludes equty provded by the angels, venture captalsts, government and/or companes. Fgure 1 descrbes the sources of fnance avalable to small busnesses over tme. In the startup stage, busnesses must rely on ntal nsder equty, angel fnance and venture captal. As startup busnesses grow over tme, they gan access to a broader spectrum of sources of external busness debt. Over tme, accumulaton of tradng hstory enables access to better sources and amounts of external fnancng, partcularly trade credt. Fgure 1 here Table 2 Panel A shows the annual balance for each source of fnancng. Consstent wth fnancal growth lfe cycle theory, the porton of owner s equty fnancng relatve to other sources of fundng s the hghest n the early years of the startup lfe cycle (53.68 n 2004). The data also ndcate that the proporton of personal debt decreases steadly over tme, whle the proporton of busness debt slghtly ncreases before the fnancal crss. However, followng the onset of the fnancal crss n 2008, the proporton of busness debt n total fnancng started to gradually declne over tme. Insder s busness debt shows a steady decrease over tme, whle both owners equty and trade fnancng balances ncreased n the same perod. Although the proporton 7

9 of nsder and outsder equty balances decreases over tme, data show that the proporton of owner s equty ncreases durng the sample perod. Ths change n balance could be due to the fact that startup busnesses are buldng up owners equty by renvestng the proft nto the busness. Fgures 2, 3, 4, and 5 here The major postulate of the fnancal growth lfe cycle s that as startups advance on the sze/age/nformaton contnuum, the fnancal sources of captal also change. The data presented n Panel B of Table 2 along wth Fgures 2 >5 show the trajectores of each source of annual captal njectons. As can be seen, the percentage of total equty njectons (owners, nsders, and outsders equty) decreased steadly from percent n 2004 to percent n On the other hand, the proporton of trade fnance as a source of fnancng ncreased from percent to per cent durng the same tme perod. Smlarly, busness debt annual njectons ncreased steadly from 7.79 percent n 2004 to percent n Meanwhle, personal debt annual njectons ncreased from percent to percent. It s worth notng that nsder s personal debt annual njectons decreased from 3.41 percent n 2004 to 2.43 per cent n 2011, whle personal credt card njectons, the major source of personal debt fnancng, ncreased from percent to per cent. In addton, the annual personal bank loan njectons decreased from 8.35 percent n 2004 to 4.48 percent n The majorty of the ncrease n busness debt njectons stems from an ncrease n the use of credt lnes from 1.33 percent n 2004 to 9.61 percent n Busness credt card njectons also show an ncrease from 1.73 percent to 7.06 percent n the same perod. Whle these descrptve statstcs help us summarze our varables n an effcent and easly understood manner, the do not confrm the sgnfcance of the trend noted n these varables over tme. Thus, we test the sgnfcance of trajectores these varables follow va a multvarate analyss. Table 2 here 3.3. Multvarate Analyss Trajectory of fnancal captal njectons The fnancal growth lfe cycle theory predcts that the major captal structure varables follow a specfc trajectory over the lfe of the busness. In ths secton, we test the postulates of ths theory, namely hypotheses H1(a) and H1(b). For the purpose of nvestgatng the systematc change (growth/declne) n the fnancal structure varables we use Latent growth modelng, employng the herarchcal lnear modelng (HLM) statstcal technque. Unlke standard regressons, the Latent growth model ncorporates the tme factor explctly. To account for nformaton opacty across the dfferent stages of the lfe cycle, we estmate the followng model: Level-1 equaton (measurement model): y t = π 0 + π 1 Tme t + π 2 Tme t 2 + ε t, for = 1,2,.. n and t = 1,2,.. T. Level-2 equatons (structural model): π 0 = γ 00 + μ 0 π 1 = γ 10 + μ 1 π 12 = γ 20 + μ 2 where y t s the response varable for frm at tme t, π 0 s a latent varable that represents the level-1 ntercept (endowments ntal status ), π 1 s a latent varable that represents the change trajectory (rate of change), and π 2 captures the acceleraton n the varables growth trajectores. 8

10 In the level-2 equatons, γ 00, γ 10, and γ 20 are the ntercepts or the average value of π 0, π 1, π 12, respectvely. The terms μ 0, μ 1, and μ 2 are random effects representng ndvdual devatons from the overall sample means. These terms dstngush each frm s ntercept (π 0 ) and slope (π 1, π 12 ) from the overall sample mean and ntercept. Heterogenety n the frm ntercept and slope parameters s determned by examnng σ 2 μ0, σ 2 μ1, and σ 2 μ02. In other words, dfferences n frm ntercepts and slopes exst f σ μ0, σ μ1, and σ μ02 are dfferent from zero. The model possesses several mportant advantages. Frst, t allows varablty [μ 1 ] n the rate of change,.e. some frms are changng more than others and have larger slopes, some are mprovng (postve slopes), some are declnng (negatve slopes), and some frms exhbt no change at all (zero slopes). Second, the model allows to ntroduce varance n both the slope and ntercept; t measures how much frms dffer at startup as well as determnes the dfference n ther 2 rates of change over tme. Thrd, Tme t allows us to control for any non-lnearty n the model. In addton to the herarchcal lnear modelng (random effect) we use the fxed effects models to estmate the trajectory of fnancal captal njectons as a robustness test. In the fxed effects models, the model to be ftted s represented by the followng set of equatons: 2 y t = π 1 Tme t + π 2 Tme t + ε t, for = 1,2,.. n and t = 1,2,.. T. ε t = μ + ν t 2 y t = μ + π 1 Tme t + π 2 Tme t + ν t In the above model, the dosyncratc error (ν t ) vares tme and among frms, where E(ν t ) = 0, Var(ν t ) 2 = σ ν, (ν t, ν jt ) = 0 for j, Cov(x t, ν s ) = 0 for all t and s and t s allowed to have Cov(μ, ν t ) 0 or Cov(μ, x t ) 0. In ths fxed effects approach, μ s treated as a frm-specfc error term n the regresson model. The frm-specfc error does not change over tme and every frm has a fxed value on ths varable. The fxed-effects model controls for all tme-nvarant dfferences among frms. Thus, we are controllng for heterogenety among frms and any other tme-nvarant unobservable varables. Table 3 here Table 3 shows the results of estmatng both the random and fxed effects models. In terms of sgnfcance, whle both models produce dentcal results the values of coeffcents they yeld are slghtly dfferent. The majorty of the fnancal captal varables exhbt a sgnfcant nonlnear relatonshp over tme. The results show strong evdence of a sgnfcant fnancal growth pattern n the use of debt and equty sources durng the lfe cycle of the frm. Consstent wth the fnancal growth lfe cycle model, the entrepreneur reles on ntal nsder fnancal captal (personal savngs, fnance from frends and famly, quas-equty, and personal debt) and trade credt. Owner equty njectons start at a hgh level and exhbt a decrease over tme. The total personal debt njectons ncrease n the frst four years of the startup lfe, after whch they start to declne. There s no evdence that outsders equty s ncreasng durng the lfe of the startup busnesses, as predcted by the fnancal growth lfe cycle model. On the other hand, busness debt and trade credt fnancng varables show a strong support for fnancal growth lfe cycle model. In partcular, both busness credt lne and busness credt card start to play an ncreasngly sgnfcant role over tme. Thus, our results strongly support the hypotheses that entrepreneurs personal resources are most mportant sources of fundng at the startup stage; over tme, the relatve mportance of nsder s equty (ncludng owners equty) n fnancng the frm declnes. The relatve mportance 9

11 of owners personal debt n fnancng the frm declnes as well, whereas the relatve mportance of busness debt n fnancng the frm ncreases over tme. The results also show that the relatve mportance of trade credt n fnancng the frm ncreases over tme Determnants of fnancal captal njectons and ts components To gan further nsght nto the factors affectng the fnancal captal njectons, we also explore the determnants of varous types of busness and personal debt njectons n total fnancal captal njecton volume for our sample of startup frms. Snce our dependent varables are ratos that are bounded by zero we use Tobt regressons to obtan consstent estmators. Let y t = the fnancal captal njectons proporton of the -th frm as a percentage of total fnancng at tme t and X t = the vector of ndependent varables of the -th frm at tme t. Snce our dependent varables are ratos that are bounded by zero, the dependent varable s truncated. 2 y 0 1x, where y 0 for 1,, N and N(0, ) Neglectng the truncaton can lead to based estmates of 0 and. 1 ( ( 0 1x ) / ) E[ y y 0, x ] 0 1x 1 ( ( 0 1x ) / ) The last term on the rght-hand-sde of the above equaton s the Mlls rato. * Latent model: y X * y 0 f y 0 y X X * f y 0 X E y y 0, X ) s gven above n the truncated regresson model. ( The uncondtonal expectaton of y gven X can be expressed as: E( y X ) = ( 1 ( X / )) X ( X / ) The total effect of a change n the k-th ndependent varable gven by: E[ y X ] x k, ( 1 ( X / )) = k x, on the expectaton of Our ndependent varables are proxes that have been wdely used n the fnance lterature to explan the determnants of captal structure n large, publcly traded frms. Ttman and Wessels (1988) argued that as larger frms become more dversfed they face lower probablty of bankruptcy. Ths lnk suggests that the larger the frm sze, the hgher the frm s debt capacty. Snce large frms tend to possess more collateralzed assets and have more stable cash flows, the company s sze s nversely related to the probablty of default allowng t to carry more debt. Damond (1991) also noted that large establshed frms have better reputaton n the debt markets, whch allows them to carry more debt. Thus the sze varable, measured by the log of total sales, s expected to be postvely related to debt fnancng. Tangble assets (collateral) convey nformaton to nvestors about the qualty of a frm as well as reduce the degree of nformaton asymmetry and opaqueness. The use of collateral s prevalent n external debt contracts as a means of mtgatng nformatonal asymmetry that may result n credt ratonng or denal of credt (Bonaccors d Patt and Dell Arcca, 2004). Typcally, small busnesses pledge assets owned by the frm,.e. nsde collateral that ncludes frm assets k y s 10

12 such as nventores, recevables, or plant and equpment. Most theoretcal research suggests that nsde collateral helps reduce adverse selecton and prevent credt ratonng (Stgltz and Wess, 1981; Besanko and Thakor, 1987a, 1987b). Asset structure s an mportant predctor of the ablty to rase ntermedated debt. Frms possessng more tangble assets are more lkely to use them as nsde collateral, therefore reducng the nformatonal asymmetry between lenders and owners. We use the rato of the property, plant, and equpment to total book assets as our proxy for collateral (Van der Wjst and Thurk, 1993; Ang, 1992). Myers (1977, 1984) suggested that frms wth dverse growth optons use less debt n order to preserve ther debt capacty. Smlarly, Rajan and Zngales (1995) and Ttman and Wessels (1988) argued that frms wth hgh growth optons depend on equty fnancng more than on debt fnancng. Ther argument suggests a negatve relatonshp between leverage and growth optons. Thus, we expect to fnd frms wth greater growth optons as measured by ntellectual property rghts and/or R&D actvty to be less relant on debt fnancng (H4). On the other hand, the fnancal growth lfe cycle model postulates that small frms wth hgh growth opportuntes are assocated wth more nformaton asymmetry, and would thus use more nternal captal and personal debt. The adverse selecton model (Myers and Majluf s (1984) mples a peckng order n fnancng small frms: retaned earnngs frst, followed by debt and then external equty. To avod the undernvestment problem owners seek to fnance new projects usng nternal funds or rskless debt. We control for nternal equty (retaned earnngs) by usng the proft to assets rato. It s expected that frms wth hgh nformaton asymmetry wll exhaust ther nternal equty before seekng external fundng. Thus, we predct a negatve relatonshp between captal njectons and proftablty. The man predcton of the lfe-cycle theory s that frms characterzed by hgh nformaton asymmetry rely more on nformal captal and less on formal credt channels. Thus, small frms frst seek nsder fnance before pursung angel fnance and trade credt and only later rely on ntermedated external fnance (Berger and Udell, 1998; Sahlman, 1990). In contrast, frms wth less nformaton asymmetry should rely on outsde debt fnancng as the man source of captal (H3). Therefore, we expect that frms wth hgh nformaton asymmetry exhbt a fnancng behavor more consstent wth the lfe-cycle theory model. We use the avalablty of credt score and fnancal dstress score as a proxy for nformaton asymmetry. In addton, n our analyses we employ the locaton of the busness (home based vs. leased locaton) as an ndcator varable whch proxes for the avalablty of nformaton about the busness. It s expected that busnesses operatng from the owner s home have a more pronounced nformaton asymmetry problem. To control for owners characterstcs we nclude gender, educaton, and race n our analyss. In addton, snce fnancal decson-makng s strctly nfluenced by the ownershp structure (Jensen and Mecklng, 1976) and to address dfferent levels of ownershp control we nclude a dummy varable for small busnesses that are owned by more than one owner. Table 4 here In Table 4 Panel A we present the coeffcents of frm and owner characterstcs that explan the relatve sze of personal debt njectons and ts components. The coeffcents for R&D are postve and sgnfcant suggestng that the hgher the growth optons, the hgher the proporton of personal debt njecton n total captal njecton volume. In addton, we show that frms wth hgher credt rsk use more nsder s personal debt and rely less on personal bank loans. The coeffcents for tangble assets show a postve and sgnfcant effect across all types of personal debt, consstent 11

13 wth pror results demonstratng that frms wth more tangble assets have hgher debt capacty. The coeffcent for sales s postve and sgnfcant for all types of personal debt njectons, except nsder s debt suggestng that frms wth a hgher level of sales have the ablty to acqure many types of personal debt. Although the coeffcent for nformaton asymmetry proxy s not sgnfcant n Panel A, suggestng that t has no mpact on the personal debt njectons, t s statstcally sgnfcant n Panel B. Thus, frms wth hgh nformaton asymmetry have lower njectons of busness debt n ther captal structure. Home based frms use less nsder s personal debt and personal bank loan njectons, but more captal njectons n the form of personal credt cards. The fndngs also suggest that proftablty lowers the need to use any type of personal debt njectons. Team owned frms tend to be less relant on nsder, outsder, and personal credt card debt njectons but they use more personal loans as captal njectons. More educated owners and owners wth longer workng experence use less personal debt njectons. In addton, male owners tend to use personal credt card less and rely more on personal bank loans. Whte owners use less nsder s debt and personal credt card njectons, whle usng more personal bank loan njectons. In Table 4 Panel B we present the coeffcents of frm and owner characterstcs that explan the relatve sze of busness debt njectons and ts components. Gven the use of busness debt fnancng, we fnd that frms wth hgh growth optons use more busness debt njectons across all types of busness debt. Moreover, frms wth ntellectual property rghts tend to have more busness debt njectons relatve to other frms. Credt rsk shows that frms wth hgher rsk wll use fewer njectons from outsders, and are also less relant on credt lne, credt card, and bank loans. Tangble assets play a sgnfcant role as collateral n case of outsder, busness credt cards, busness bank loans, and other busness loan njectons. Large frms tend to use more busness debt njectons. The coeffcent for our nformaton asymmetry proxy s negatve and sgnfcant, suggestng that nformatonally opaque frms are less lkely to use outsders debt, credt lne, credt card, and bank loan captal njectons. Ths result s consstent wth the theoretcal predctons of Stgltz and Wess (1981) and Berger and Udell (1998) who noted that nformatonally opaque frms are less lkely to access debt fnancng due to severe frctons n the debt markets. Our results also show that home based frms are less lkely to use busness debt njectons across all types of busness debt avalable to the frms. Proftable, less constraned frms use fewer busness debt njectons. We also show that several owner characterstcs explan the proporton of busness debt njectons n startup frms. For example, gender s a sgnfcant factor n determnng the proporton of busness debt njectons. Across all types of busness debt njectons, male owners use more busness debt relatve to female owners. Wth respect to race, whte entrepreneurs rely less on nsder s busness njecton and credt card fnancng. On the other hand, we fnd that owners' level of educaton and work experence do not affect the proporton of busness debt njectons. In Table 4 Panel C we present the coeffcents of frm and owner characterstcs that explan the relatve sze of equty njectons and ts components. Hgher growth optons ncrease both the proporton of equty njectons and trade fnance. Frms wth hgh credt rsk rely more on owner s equty njectons and less on debt njectons or trade fnance njectons. Frms wth hgher sales use fewer owner s equty njectons and rely more on debt or trade fnance njectons. The results also show that hgher nformaton asymmetry prompts frms to use more equty njectons and less trade fnance or debt njectons. Home-based frms rely on usng owner s equty more and nsders and outsders equty njectons less. Proftablty reduces the use of both equty and debt captal njectons. Team owned frms use more equty njectons and less debt njectons. Male owners use less owner s equty njectons and rely more on outsders equty and trade fnance captal njectons. Fnally, whte owners use less equty njectons whle favorng trade fnance njectons. 12

14 4. Conclusons In ths artcle we examne small busness fnancng through a growth cycle paradgm n whch dfferent combnatons of captal are optmal at dfferent ponts n the busness lfe cycle. Unlke publcly held frms, small busnesses n the startup stage fnance ther operatons wth nsders captal such as owners equty, funds from famly and frends. As the busness grows, outsders sources of captal may become avalable. The key feature that drves the fnancal growth cycle s nformatonal opacty. Early n the growth cycle, small busnesses typcally lack busness assets that can be used as collateral, have lttle repayment hstory or record of proftablty, and do not have audted fnancal statements. Therefore, provders of outsde captal put consderable weght on the credtworthness and reputaton of the entrepreneur who may have a longer credt hstory and personal assets that can be used as collateral. Our results pont to a sgnfcant fnancal growth pattern n the use of varous sources of debt and equty durng the lfe cycle of the frm. Consstent wth Berger and Udell s model, at the startup stage entrepreneurs rely on ntal nsder s captal sources. Over tme, as frms become less opaque, the proporton of busness debt and trade credt fnancng n total captal njectons ncreases sgnfcantly. Although the proporton of owner s equty n total captal njectons decreases over tme, the annual balance of owner s equty ncreases suggestng that owners use retaned earnngs to ncrease ther ownershp stake n the frm. Several busness and owner characterstcs explan the type and magntude of varous types of captal njectons. Frms characterzed by hgh nformaton asymmetry rely more on equty njectons, especally owner s personal savngs, whereas those undertakng nnovatve actvtes fnance ther operatons wth outsders equty njectons. A hgher proporton of tangble assets explans greater use of debt njectons (both personal and busness debt sources), whereas hgher level of sales drves frms to use both debt njectons and trade fnance, relyng less on equty njectons. Frms owned by two or more owners tend to nject more equty (especally outsde equty) and less debt n the captal structure of ther frms. Educaton and race have a sgnfcant mpact on the captal njectons over the busness lfe cycle. Hghly educated owners choose to nject lower proportons of personal debt and trade fnancng, whereas whte owners nject lower proportons of personal equty and rely more on trade fnancng. Ths artcle contrbutes to the current stream of lterature by provdng nsghts nto the ways n whch the sources and magntude of captal njectons n startup frms dffer from those of large, publcly held frms. The reported fndngs reveal the systematc changes n the fnancal structure of newly formed busnesses and show that there s a sgnfcant fnancal growth pattern n the use of varous sources of debt, equty, and trade fnance durng the lfe cycle of small frms. Fnally, our research results demonstrate that entrepreneurs personal characterstcs play an mportant role n the use and magntude of a specfc type of captal njecton. 13

15 References Ang, J. S., On The Theory of Fnance for Prvately Held Frms. Journal of Small Busness Fnance 11 (3), Ballou, J., Barton, T., DesRoches, D., Potter, F., Zhao, Z., Santos, B., Sebastan, J., Kauffman Frm Survey (KFS) Baselne Methodology Report. Avalable at: Baxter, N.D., Leverage, Rsk of Run and the Cost of Captal. Journal of Fnance 22, Berger, A. N., and Udell, G.F., The Economcs of Small Busness Fnance: The Roles of Prvate Equty And Debt Markets In The Fnancal Growth Cycle. Journal of Bankng and Fnance 22, Bertrand, M., Schoar, A., Managng Wth Style: The Effect of Managers on Frm Polces. Quarterly Journal of Economcs 118, Besanko, D., Thakor, A.V., 1987a. Collateral and Ratonng: Sortng Equlbra n Monopolstc and Compettve Credt Markets. Internatonal Economc Revew 28, Besanko, D., Thakor, A.V., 1987b. Compettve Equlbrum n the Credt Market under Asymmetrc Informaton. Journal of Economc Theory 42, Bonaccors d Patt, E. and Dell Arcca, G., Bank Competton and Frm Creaton. Journal of Money Credt and Bankng, 36 (2), Cassar, G., The Fnancng of Busness Start-Ups. Journal of Busness Venturng 19, Coleman, S. and Robb, A Captal Structure Theory and New Technology Frms: Is There a Match? Management Research Revew 35, Coleman, S., Cote, C., and Farhat, J The Debt-Equty Fnancng Decsons of U.S. Startup Frms. Journal of Economcs and Fnance, Volume 40, Cole, R., Goldberg, L.G and Whte, L.J Cooke-Cutter versus Character: The Mcro Structure of Small-Busness Lendng By Large and Small Banks. Journal of Fnancal and Quanttatve Analyss 39, Cole, R. and Sokolyk, T How Do Start-Up Frms Fnance Ther Assets? Evdence from the Kauffman Frm Surveys. Avalable at SSRN: or Damond, D., Montorng and Reputaton: The Choce between Bank Loans and Drectly 14

16 Placed Debt. Journal of Poltcal Economy 99 (4), DeAngelo H. and Masuls, R., Optmal Captal Structure under Corporate and Personal Taxaton. Journal of Fnancal Economcs 8, Easterbrook F. H., Two Agency-Cost Explanatons of Dvdends. Amercan Economc Revew 74, Farhat, J. (2012). Usng Weghts n the Analyss of the KFS Data: KFS Professonal Development Workshops. Ewng Maron Kauffman Foundaton Research Paper. Avalable at SSRN: Farhat, J. and Robb, A. (2013). Analyzng the KFS Multply Imputed Data.Avalable at SSRN: Farhat, J. and Robb, A. (2014). Appled Survey Data Analyss usng STATA: The Kauffman Frm Survey Data, Ewng Maron Kauffman Foundaton Harrs, M., and Ravv, A Captal Structure and the Informatonal Role of Debt. Journal of Fnance 45, Hart, O., and Moore, J Debt and Senorty: An Analyss of the Role of Hard Clams n Constranng Management. Amercan Economc Revew 85, Jensen M. and Mecklng, W Theory of the Frm: Manageral Behavor, Agency Costs, and Captal Structure. Journal of Fnancal Economcs 3, Kraus, A., and Ltzenberger, R A State-Preference Model of Optmal Fnancal Leverage. Journal of Fnance 28, McInsh, T.H., Indvdual Investors and Rsk-Takng. Journal of Economc Psychology 2, Myers, S.C., Determnants of Corporate Borrowngs. Journal of Fnancal Economcs 5, Myers, S.C., The Captal Structure Puzzle. Journal of Fnance 39, Myers, S.C. and Majluf, N.S., Corporate Fnancng And Investment Decsons When Frms Have Informaton That Investors Do Not Have. Journal of Fnancal Economcs 13, Nofsnger, J. R. and Wang, W., Determnants of Start-Up Frm External Fnancng Worldwde. Journal of Bankng and Fnance 35, Rajan, R. and Zngales, L What Do We Know About Captal Structure? Some Evdence from Internatonal Data. Journal of Fnance, 50,

17 Robb, A. and Robnson, D. T The Captal Structure Decsons of New Frms. The Revew of Fnancal Studes 27, Robchek, A., and Myers, S Problems n the Theory of Optmal Captal Structure. Journal of Fnancal and Quanttatve Analyss, June ssue, Sahlman, W The Structure and Governance of Venture-Captal Organzatons. Journal of Fnancal Economcs 27, Scott, J., A Theory of Optmal Captal Structure, Bell Journal of Economcs and Management Scence, Sprng ssue, Stgltz, J., and Wess, A Credt Ratonng In Markets wth Imperfect Informaton. Amercan Economc Revew 71, Stulz, Rene, Manageral Dscreton and Optmal Fnancng Polces. Journal of Fnancal Economcs 26, Ttman, S. and Wessel, R The Determnants of Captal Structure Choce. Journal of Fnance 43, Verheul, I. and Thurk, R Start-Up Captal: Does Gender Matter? Small Busness Economcs 16, Van der Wjst, N. and Thurk, R Determnants of Small Frm Debt Ratos: An Analyss of Retal Panel Data. Small Busness Economcs 5,