When trade credit facilitates access to bank finance: Evidence from US small business data. Version : February 2006

Transcription

1 When trade credit facilitates access to bank finance: Evidence from US small business data Pascal Alphonse Jacqueline Ducret Eric Séverin Version : February 2006 Abstract: While trade credit is traditionally considered as a substitute for bank loans, recent theoretical papers (e.g. Biais and Gollier (1997), Buckart and Ellingsen (2004)) suggest that bank debt and trade credit can also be considered as two complementary sources of financing. We suggest that these two hypotheses are related to different classes of firms and provide a missing link between them based on the incentive for good firms to invest in reputation. By using US small businesses data (NSSBF 1998), this paper provides an empirical analysis of these hypotheses. The empirical findings are consistent with the hypothesis that trade credit helps firms to improve their reputation. The results show that trade credit can work as a signal about firm s quality and thus facilitates access to bank debt. JEL Classification : G32 Key words : bank Debt, Small Business Finance, Trade Credit. Acknowledgements The authors wish to thank the financial support of the Commissariat Général au Plan, the Board of Director of the Federal Reserve of the availability of the data as well as participants at the AFFI (Lyon), EFMA (Basel) and MFS (Istanbul) meetings. IAE, Université de Nancy rue Michel Ney, Nancy. pascal.alphonse@univnancy2.fr (corresponding author) IAE, University of Valenciennes, Rue des cent Têtes, Valenciennes cedex Jacqueline.Ducret@univvalenciennes.fr IAE, University of Lille 1,USTL, 104, avenue du people belge, Lille cedex. eric.severin@iae.univlille1.fr 1

2 When trade credit facilitates access to bank finance: Evidence from US small business data Abstract: While trade credit is traditionally considered as a substitute for bank loans, recent theoretical papers (e.g. Biais and Gollier (1997), Buckart and Ellingsen (2004)) suggest that bank debt and trade credit can also be considered as two complementary sources of financing. We suggest that these two hypotheses are related to different classes of firms and provide a missing link between them based on the incentive for good firms to invest in reputation. By using US small businesses data (NSSBF 1998), this paper provides an empirical analysis of these hypotheses. The empirical findings are consistent with the hypothesis that trade credit helps firms to improve their reputation. The results show that trade credit can work as a signal about firm s quality and thus facilitates access to bank debt. 2

3 Introduction In recent years, the banking industry has experimented important modifications and restructurings and has also faced regulatory (the single market program in the European Union, the GrammLeachBliley Act in the US, the Basel II agreement) and technological changes that might affect the aggregate amount of credit supplied to the economy as well has the composition of banks credit portfolios. Small businesses may be particularly affected by these changes because of their dependency on financial institutions for external finance. For example, Berger and Udell (1995) highlight that large banking institutions devote lesser proportions of their assets to Small and Medium Enterprises (SME) than small banking institutions, suggesting that consolidation in the financial system reduce the availability of credit to SME 1. Small business lending is also affected by the monetary policy. Following the bank lending view, one can expect a tightening monetary policy to decrease the loan supply from banks. This credit rationing appears particularly harmful for the SME (Kashyap and Stein, 1997) which may find it impossible to finance their growth and their investment (Bernanke et al. 1996). Besides, the tightening of monetary policy has implications on interest rates. High interest rates impair the collateral value and reduce the net worth of SME making them less creditworthy and decreasing their ability to raise funds, a phenomenon known as the financial accelerator mechanism. Indeed, the lack of available funds accelerates the negative effects of macroeconomic shocks which, in turn, worsens SME financial situation which is very dependent of bank loans. In front of the difficulty to raise funds, SME are compelled to reduce (at least for a part and in totality in the worst cases) their investments. This situation exacerbate regional and macroeconomic difficulties (Bernanke, 1983). 1 Financial distress in the banking industry may also impact the lending to SME. Berger and Udell (1998) show that in order to reduce their risk exposure and to retrieve their capital ratios, banks can ration SME credit and develop a politic of arm s length debt. Regulatory changes also influence the supply of loans. For example, the Mc Donough ratio determined by the Basle Accord compels banks to raise capital requirements against business loans (Berger et Udell (1994)). 3

4 In order to alleviate this funding problem, SME may try to raise funds from other nonfinancial creditors, namely the suppliers. Indeed, suppliers have incentives in offering credit to their customers. According to Petersen and Rajan (1997, p. 689), the supplier has an implicit equity stake in the firm equal to the present value of the margins he makes on current and future sales of the product to the firm. This may far exceed the implicit equity stake a financial institution may have because of the potential for future business, and may explain why suspect growing firms tend to be financed by suppliers. Moreover, suppliers may have abilities to circumcise the traditional problems of informational asymmetry and moral hazard at least as well, if not better than banks. There is at least three reasons explaining this advantage of suppliers visàvis of banks. The supplier is supposed to possess a better knowledge of the technology and of the markets of its customers and hence can appraise their quality with a greater precision than banks do. A supplier may also threaten to stop future supplies and may be in a better position to repossess and resell goods in case of default than banks. (Mian and Smith (1992)). Finally, in lending goods, not cash as banks do, suppliers are less concerned with cash diversion by their customers (Buckart and Ellingsen (2005)). Building on the previous arguments (or variations of some of them), it is traditionally argued that when bank debt is unavailable for some firms, these firms can alleviate their credit constraint in raising the amount of trade credit extended by their suppliers. In that case trade credit and bank debt are considered as (possibly imperfect) substitute. Biais and Gollier (1997) go one step further and suggest that trade credit may also act as a signal of a firm s quality. More specifically, they show that trade credit reveals the private information of the supplier to the bank which, in turn, can update its beliefs about customer default risk. It can thus mitigate the information asymmetry which otherwise would prevent the financing of some positive NPV investment opportunities. Buckart and Ellingsen (2005) reach the same kind of conclusion, showing that for some firms, the supply of trade credit increases the bank credit limit, hence making trade credit and bank finance complements. Hence, following the theoretical literature, trade credit could be considered either as a substitute or as a complement to bank debt and the primarily goal of the present study is to discriminate between the two hypotheses. Nevertheless, the theoretical models quoted above are essentially static and we may think that only the less established firms with no reputation 4

5 have to deal with severe asymmetric information or moral hazard problems. Put another way, we propose to link the two seemingly competing effects (complemetarity vs. substitutability) to different classes of small businesses. More precisely, we argue that there may be a dynamic element that articulates the two hypotheses and devise a test of this new hypothesis on US small business data taken from the 1998 NSSBF survey. There is only a few papers that address the complementary role of trade credit vs the substitution hypothesis. Petersen and Rajan (1997) implement comprehensive tests based on small business data and provide a supply and demand analysis of trade credit among SME which mainly support the substitution hypothesis. Nearest to our work, the empirical study of Cook (1997) is to the best of our knowledge the only one that addresses this important signalling role of trade credit. Our paper differs from hers in several aspects. Firstly, the author uses data from Russia, suggesting that trade credit provides a signal that leads to more bank credit because the Russian banking system is less developed and less efficient from the informational point of view than the US one 2. In contrast, our results rely on the US financial architecture and banking system. Secondly, her methodology is based on a probit model which postulates that the direction of causality runs from trade finance to bank finance. In that case, trade credit is considered as exogenous which is at odds with traditional findings (e.g., Petersen and Raja, 1997) On the contrary, we take into account the possibility of trade finance being endogenous, using a simultaneous equations system. Our results suggest three main conclusions : Firstly, using recent US small business data, we provide new evidence concerning the role of trade credit as a substitute for bank debt. Indeed, an increase in bank debt lowers the amount of trade credit the firm uses This is consistent with Petersen and Rajan (1997). When firms are credit constrained by banks, they have to use trade credit. Secondly, and new in this paper, we show that trade credit is a complementary source of financing to bank debt. This result is consistent with Biais and Gollier (1997) and Buckart and Ellingsen (2004). 2 See also DemirgucKunt and Maksimovic (2002) who point out that firms use of bank debt relative to trade credit is higher in countries with efficient legal system. 5

6 Thirdly, our results highlight that this complementarity effect does not hold for firms with a long banking relationship but only for firms with a short banking relationship. This is consistent with the view that trade credit could help to enhance firm reputation towards banks. The remainder of the paper is organized as follows. Section I develops the literature on trade credit. Section II describes the data and presents the methodology. Section III reports the results. Section IV offers conclusions and discusses the implications for future research. I. Related literature and hypotheses Trade credit credit extended by a seller who does not require immediate payment for delivery of a product is a very old practice, dating back to the Middle Ages (Le Goff, 1957). It is also an important source of funds for business customers. For example, Elliehausen and Wolken (1993) report that trade credit represented about 20% of all nonbank nonfarm US businesses liabilities, and up to 35% of their total assets in Rajan and Zingales (1995) calculated that trade credit represented 17.8% of total assets for all American firms in 1991, and in European countries such as Germany, France and Italy, trade credit represents more than a quarter of total corporate assets. Furthermore, trade credit is related to macroeconomics and monetary policy (see the classical papers of Meltzer (1960), Bernanke (1988) and Ramey (1993) and recent papers by Nielsen (1999) and Mateut and al. (2003)). The economic significance of trade credit as a source of funds in diverse economic conditions raises questions about its use by firms. Why do firms use trade credit and when? How should we explain the crosssectional variation in the use of trade credit? At first glance, trade credit can be a convenient cash management tool, allowing a better match of timing between cash inflows and outflows. In this vein, Ferris (1981) argues that trade credit may reduce the transaction costs associated with cash management in allowing firms to cumulate obligations and pay them on specific dates. By using trade credit, a firm can also reduce the effect of growth on working capital needs (on this aspect traditional see corporate finance textbooks, (see, for example, Damodaran, 2001 ; Brealey and Myers, 2002). This suggests that the use of trade credit may be related to a firm s revenues and current asset items such as inventories and account receivables, which are themselves related 6

7 to a firm s field of activity or industry (Emery, 1987) 3. Crosssectional variations in the use of trade credit related to industry dummies and industry growth rate dummies are reported in Fishman and Love (2001). These industry effects may also be related to moral hazard problems between a business customer and its supplier. In that case, trade credit is used as a device to manage and control the quality of the items sold by the supplier, allowing the business customer to verify product quality before paying (Smith (1987) and Long, Malitz and Ravid (1993)). This quality concern depends on product complexity and industry as well. Burkart and Ellingsen (2005) also built a agency model of the use of trade credit on the fact that it is less profitable for an opportunistic borrowers to divert input than to divert cash. This difference gives suppliers a monitoring advantage over banks that is found in the input transaction itself. Hence input illiquidity facilitates trade credit. Suppliers can also offer trade credit to business customers in order to price discriminate (Schwartz and Whitcomb, 1979 ; Brennan et al., 1988 ; Mian and Smith, 1992). In that case trade credit provides sellers with unique information about their business customers, and especially their credit worthiness (Danielson and Scott, 2000, 2002). More generally a supplier may benefit from its knowledge of industry specificities, from the history of recurrent transactions with its customers (which is, at least in part, private information) and from to the jointproduct nature of trade credit put forward in Buckart and Ellingsen (2005) when he decide to extend trade credit to one of its customers. Hence, suppliers may possess some advantages over traditional financial intermediaries in collecting information on other nonfinancial firms, in assessing their credit worthiness and finally in controlling their actions. As a consequence, one can expect the supplier to be able to discriminate between good and bad firms in troubled periods and thus to provide some of these firms in financial distress with financial support better than banks can do (Dietsch and Lartisien, 1994 ; Wilner, 2000). Petersen and Rajan (1997) suggest that this attitude of the supplier may not be restricted to firms in financial distress and should be more general. They argue that a supplier "may want to protect the value of its implicit equity stake in the customer the present value of the margin he makes on current and futures sales of the product to the firm by providing it with 3 See also the discussion in Petersen and Rajan (1997), pp

8 shortterm financing" and conclude their paper in noting that "it may explain why suspect growing firms tend to be financed by suppliers" 4. All in all, the previous discussion suggests that suppliers may be in a better position than traditional financial intermediaries to lend to their business customers, especially when the latter are small, young and opaque firms (Berger and Udell (1995)) 5 or operate in countries with poorly developed financial institutions (Fishman and Love (2001)). Therefore, it may be the case that suppliers could sometimes be able to lend when banks cannot. Thus, equilibrium credit rationing related to exante information asymmetry (Stiglitz and Weiss, 1981)) may result in more use of trade credit 6. This analysis suggests that bank debt and trade credit are two (somehow imperfect) substitutable financial resources and is therefore referred to as the substitution hypothesis. While the empirical evidence generally supports this view (see, for example, Petersen and Rajan, 1997) 7, we should observe that in doing so, bank debt is traditionally supposed exogenous or predetermined. By contrast, modern finance theory views all financing claims as inputs in a portfolio or pool of funds approach. At least, we should control for some kind of simultaneity in bank and trade credit finance. That s the reason why we provide here a new examination of this substitution hypothesis, taking into account the possibility that bank and trade credit are simultaneous determined. One positive reason for this simultaneity is provided by Biais and Gollier (1997) who challenge the traditional substitution hypothesis. They argue that the use of trade credit can alleviate the credit constraint some firms suffer from due to imperfect information and credit rationing, both directly (this is the traditional view where trade credit acts as a substitutable source of funds) and, most importantly for our purpose, indirectly (banks agree to lend when suppliers also lend to their business customers). Therefore, the use of trade credit acts as a 4 Petersen and Rajan (1997), pp See also Danielson and Scott (2000) who extend these results and provide a more indepth analysis of the role played by the traded credit discounts.. 5 More precisely, Berger and Udell (1995, pp. 360) point out age reflects public information whereas the length of relationship reflects private information available only to the lender, and corresponds to the difference between information obtained as a result of reputation versus information obtained from monitoring. 6 Schwartz (1974) is traditionally considered as the first paper pointing out this aspect. 7 In a recent paper, Mateut and Mizen (2002) show that suppliers evaluate the creditworthiness of firms on much the same basis as banks, with solvency, credit risk and age all improving the access to trade credit. They conclude that trade credit is taken up by firms as a substitute for bank finance at the margin when they are credit constrained. 8

9 signal and reveals supplier s unique information to the bank. In their model, credit rationing occurs, in a first round, because the bank cannot always assess the quality of a firm with enough precision to make the types of loans banks make. As a consequence, some firms with positive net present value projects cannot be financed with bank debt. Nevertheless, suppliers can sometimes find it profitable to finance some of these firms and then extend trade credit to them. In a second round, banks can observe this actual use of trade credit and update their beliefs concerning the quality of firms. When the equilibrium of the game is reached, some firms which would have suffered from credit rationing in the absence of trade credit finance actually finance positive net present value projects with a mix of trade credit and bank debt. This view makes trade credit and bank debt two complementary financial resources and is therefore referred to as the complementary hypothesis 8. Buckart and Ellingsen (2005) reaches a somehow similar result but rooted it in their input transaction model. More specifically, in their model, additional trade credit increase the investment size and thereby the entrepreneur s residual return and hence decreases the entrepreneur s incentive to divert cash. As a consequence bank debt credit limit increases, making bank debt and trade credit complements. Note that in these models the two effects are simultaneous: firms use trade credit because they are credit constrained, this is the substitution effect but trade credit also facilitates access to bank debt, this is the complementary effect. Unfortunately, they may not occur at the same point in time and thus have to be disentangled in empirical assessments. In this paper we suggest that firms may use trade credit in order to attempt to built some (good!) reputation in the borrowing market. More specifically, we argue that firms that benefit from long term banking relationship have no incentives to use trade credit 9 : for these firms only the substitution hypothesis should be relevant. For firms (say a sample of) with poor banking relationship, the substitution hypothesis and the complementary hypothesis should both be relevant because some firms (of good quality) use trade credit to signal themselves (this is the 8 For recent assessments on the nature of trade credit visàvis bank debt, see Uesugy and Yamashiro (2004), Bond (2005), Berlin (2004), Buckart, Ellingsen and Giannetti (2005). 9 Indeed, since the early work of Rajan (1992), financial economists have reached a consensus on the duration s and intensity banking relationship influence on availability of credit lines. Whatever the indicators chosen and the countries, all the studies highlight that duration and intensity of relationship banking allow firms to access better access to bank credit (Petersen and Rajan, 1994 for USA; Elsas and Krahnen, 1998 for Germany ; d Auria et al., 1999 for Italy). 9

10 complementary hypothesis put forward in Biais and Gollier (1997) and, to some extent, by Buckart and Ellingsten (2005)) while other firms (of bad quality) use trade credit only because they cannot do otherwise. We refer to this hypothesis as the reputationnal hypothesis. II. Data and methodology A. Data The data in this study have been obtained from the National Survey of Small Business Finance (NSSBF) which contained detailed crosssectional information on small firms. The survey was conducted in under the guidance of the Board of Governors of the Federal Reserve System. It targeted forprofit, nonfinancial, nonfarm, nonsubsidiary businesses with fewer than 500 workers which were in operation as of December, The survey collected information on the availability and use of credit and other financial services by small businesses along with information on firm characteristics including number of workers, number of owners, organizational form, location, credit history, income statement and balance sheet data. The survey is regarded as the most comprehensive and detailed source of crosssectional information on small business finance 11. The data from previous surveys dating from 1987 and 1993 have been extensively used in financial research and noticeably in some of the most important papers about banking relationships (see, for example, Petersen and Rajan, 1994 ; and Berger and Udell, 1995). There are 3561 firms in the sample. The sample can be considered as representative of the 5.3 million American small businesses. It is stratified by region, rural/urban location, 10 According to US SBA, Office of Advocacy, Small Businesses account for half of the private sector output, employ more than half of privatesector workers, and provide about threequarters of net new jobs each year ( 11 See Wolken (1998) for an analysis of the merits of the various data sets concerning small businesses. 10

11 ethnic owner type and size to ensure that rural, minorityowned and large businesses are represented in the sample. We chose not to take into account the smallest firms, the microenterprises, for which we suspect very specific behaviour (Berger and Udell, 1998) 12. Therefore, we only selected Small and Medium Entreprises (SME), that is, firms that employ more than 20 workers and less than 500. We excluded firms operating in the mining and financial sectors, as well as firms for which the business field was not mentioned. Finally, firms with too many missing or inconsistent data 13 were discarded. The final sample comprises 654 firms 14. A first look at the data indicates that 91% of these firms are corporations (either S corporations or C corporations) and 69% of these firms are family businesses. Sample median value for the number of workers is 62 people and for corporate age, it is about 18 years old. It amounts to 6 million dollars for firms sales and to 2.4 million dollars for firms book value. Descriptive statistics about main balancesheet items are presented in Table I. <Insert Table I about here> 12 For example, Woodruff (2001) notes that the microenterprises have very little access to formal (bank) loans, and informal credit loans from family members or friends is much more common among these very small firms. This would certainty impact the relationship between bank loan and trade credit, and therefore would make our analysis incomplete for these very small firms. Furthermore, we would have to take into account variables related to entrepreneurship and entrepreneurial choice, dealing with very small firms (Bitler et al., 2001 ; Evans and Jovanovic, 1989). Such analysis may complement the present work and is left on the agenda for further research. 13 For example, when total assets were negative or different from total liabilities. 14 The SAS code for the selection of the data is available upon request from the author. 11

12 For the sample considered as a whole, trade credit represents about 12% of total assets 15, while equity represents 51% and loans 27%. Moreover 85% of the firms use trade credit. It can then be considered as a major source of financing. Rather surprisingly, there seems to be no clear difference between firms financial structure depending on their size. On the contrary, age seems to be a relevant criteria to characterize firms financial structure as young firms are less capitalized and more indebted than old ones. The link between age and financial structure could result from the fact that among young firms, there may be high growth firms that have to rely on debt to face their important financial needs. This conjecture is corroborated by the fact that firms total assets are rather high for firms younger than 5 years, compared to older ones and particularly firms aged between 6 and 20. Nevertheless, as far as accounts payable are concerned, age does not seem to make much difference. 15 This can be compared with the results of Mateut and Mizen (2002) who reported trade credit represent more than 17% of total assets for US firms and more than 25% for French, German and Italian firms. 12

13 B. Methodology We use a simultaneous equations model to capture some key features concerning the twoway relationship between trade credit and bank debt. There are two equations in this system. The first equation is related to the use of trade credit according to the substitution hypothesis. It supposes a negative link between the amount of accounts payables and the amount of total debt. The second equation is related to the use of bank debt by firms according to the complementary hypothesis. It supposes a positive link between the amount of bank debt and the amount of account payables. The model is estimated using the two stages least squares method 16. More precisely, the first equation is identified as follows : ACCP = f(totdebt; NULINE; CURRASS; SALESGR; INCOME; SIZE; AGE; PROFIT; OWNER; MGRDUM; INDUSDUM; LEGALDUM) With : ACCP: accounts payables / total assets TOTDEBT: total loans / total assets NULINE: unused lines of credit / total assets CURRASS: current assets / total assets SALESGR: sales growth between 1997 and 1998 INCOME: log of total income SIZE: log of the book value of assets AGE: log of (1+firm age) PROFIT: earnings before tax / total assets OWNER: ownership share of the principal owner 16 Two stages least squares methods provide estimators that are consistent and efficient and less subject to specification errors than full information methods (Greene, 2003). 13

14 MGRDUM: =1 if the firm manager is a hired employee (=0 otherwise) INDUSDUM: firm industry dummies LEGALDUM: firm legal form dummies It attempts to explain cross sectional variations in the use of trade credit by firms as measured by the account payables over total assets ratio (ACCP). According to the substitution hypothesis, this ratio should be inversely related to the amount of debt defined as total loans divided by total assets (TOTDEBT) 17. We also measure credit availability to firms with the ratio of the unused lines of credit over total assets (NULINE) for which we expect a negative impact on trade credit use. This equation also includes variables related to the transaction use of trade credit. These variables are the ratio of current assets over total assets (CURRASS), a measurement of the sales growth between 1997 and 1998 (SALESGR) and the (log of) total income (INCOME). We expect these variables to be positively related to the amount of trade credit used. Trade credit may be viewed as a means of challenging problems of moral hazard and asymmetric information which are related to firm opacity. Following Berger and Udell (1998, 2002), the size of the firm is chosen as a proxy for firm opacity (see, for example, Fluck, 1999) 18. The size of the firm, is measured by the (log of) the book value of assets (SIZE). We expect this variable to be negatively related to the amount of trade credit that firms use. Firm size is not the only indicator which may reveal the degree of opacity of the firm. Firm age may also be considered as an indicator of asymmetry of information. But, it seems 17 Note that we include in this ratio all the loans to the firm, including loans from the shareholders. The reason is that with this variable we try to capture the possible effects of capital constraint on trade credit use. 18 Of course, firm size may be related to many other concerns than financial opacity. Beck et al., (2003) and Kumar et al., (2001) provide several interpretations of the determinants of firm size. 14

15 that firm age may be linked to many others aspects, such as the amount of retained earnings and the financial resources needed, for example. Even if the notion that the source of financing evolves with the growth of the firm is well established in the literature (Berger and Udell, 1998), it is very difficult to determine the precise way firm age influences the use of trade credit. Thus, we use the log of (1 + firm age) and the square of this variable to account for the possibility of nonlinearity and opposite behaviours at different ages. According to the financial pecking order theory, the more retained earnings a firm has accumulated, the less it should be dependant on external finance, and especially trade credit 19. Thus, we take into account firm s profitability (measured by the ratio of earnings before tax over total assets PROFIT) in the equation. Profitability is certainly not a very good measure of retained earnings, but it captures at least some aspects of the implied process. The data doesn t make it possible for us to get a better proxy for retained earnings. The ownership share of the principal owner (OWNER) is further included because it may be related to different financing patterns between growth and nongrowth oriented small business (Holmes and Zimmer, 1994). Furthermore a dummy variable (MGRDUM) is included to separate ownermanaged businesses from businesses where the manager is a hired employee. Finally industry dummies (INDUSDUM) and firm legal form dummies (LEGALDUM) are introduced in the equation (Ng et al., 1999). The second equation is identified as follows : BKDEBT = g(accp; BKREL; SIZE; AGE; DEPRASS; STOCKDUM; LEASEDUM; MSADUM; SALESGR; PROFIT; OWNER; MGRDUM; INDUSDUM; LEGALDUM) With : BDEBT: bank debt / total assets 19 The literature on small business finance stresses the ability of the pecking order approach in explaining the capital structure of small firms. See, for example, Romano, et al., (2000) or Chittenden, et al., (1996). 15

16 ACCP: accounts payables / total assets BKREL: log of (1+ the longest duration of the relationship with a bank) SIZE: log of the book value of assets AGE: log of (1+firm age) DEPRASS: depreciable assets / total assets STOCKDUM: =1 if the firm benefits from loans from stockholders (=0 otherwise) LEASEDUM: =1 if the firm uses capital lease (=0 otherwise) MSADUM: =1 if the firm is located in urban area (=0 otherwise) SALESGR: sales growth between 1997 and 1998 PROFIT: earnings before tax / total assets OWNER: ownership share of the principal owner MGRDUM: =1 if the firm manager is a hired employee (=0 otherwise) INDUSDUM: firm industry dummies LEGALDUM: firm legal form dummies It tries to explain cross sectional variations in the use of bank debt as measured by the amount of bank debt over total asset ratio (BKDEBT) by firms. According to the complementary hypothesis, trade credit use should positively influence the level of bank debt used by firm. Therefore the ratio of accounts payable over total assets (ACCP) is included in this equation and we expect a positive impact on bank debt use. The duration of the banking relationship is also included in the form of the log of one plus the longest duration of the relationship with a bank (BKREL) because it has been suggested that firms benefit from these relationship essentially in term of credit availability 16

17 (Petersen and Rajan, 1994). We take the square value of this variable as well as the variable itself to take into account possible nonlinearity in the relation. The use of bank debt should be positively related to firm size (SIZE) as an inverse indicator of firm opacity. Firm age (AGE) is also introduced into the equation. Considering firm age may be linked to firm opacity, financial needs, it is difficult to predict the precise way it may influence the use of bank debt 20. The use of bank debt should be positively related to tangible assets that might be pledged as collateral, thus the depreciable assets over total assets ratio is inserted in this equation (DEPRASS). Three dummy variables are introduced as indicators for fund availability and financial needs. The first dummy variable takes the value of one if the firm benefits from loans from stockholders and zero otherwise (STOCKDUM). The second dummy variable takes the value of one if the firm uses capital lease, and zero otherwise (LEASEDUM). We expect these variables to be positively associated with bank debt. Finally, a third dummy variable is related to the localisation of the firm in urban vs. rural area (MSADUM). Sales growth might be positively related to financial needs and is therefore introduced in the equation too (SALESGR). Profitability (PROFIT) as an indicator of financial needs is also introduced in the equation. Finally, the two variables related to the control of the firm are included ownership share of principal owner (OWNER) and a dummy variable for owner or hired employee management (MGRDUM) as well as industry (INDUSDUM) and legal forms (LEGALDUM) dummies. 20 We use the log of (1+firm age) and the square of this variable to account for the possibility of non linearity. But, the square variable appears non significant and was eliminated from the equation. 17

18 Note that using crosssectional data, the tests of the complementarity and the substitutability hypotheses are not mutually exclusives. In fact we may observe both complementarity and substitutability effects if the coefficient of total loans divided by total assets (TOTDEBT) in the trade credit equation is negative and significant and if the coefficient of the ratio of accounts payable over total assets (ACCP) is simultaneously positive and significant in the debt equation. As we suggest in the previous section, this phenomenon may be related to the pooling of different classes of firms in the same sample. Therefore, in order to test the reputationnal hypothesis developed above, we have to split the original sample according to a relational banking intensity dummy. We expect the coefficient of the ratio of accounts payable over total assets (ACCP) to be insignificantly different from zero in the debt equation for the firms with the stronger relational banking intensity whereas we expect a positive and significant coefficient for the firms with the poorer relational banking intensity. III. Results A summary of the data is reported in Table II. <Insert Table II about here> In Table III, we report the results of the simultaneous equations model 21. <Insert Table III about here> First, we can notice that adjusted R2 are similar to those of previous studies (Biais et al., 1995). This point is rather positive since NSSBF data did not make it possible to take into 21 Results for firm s industry and legal structure dummies are not reported. 18

19 account nondebt tax shields or growth opportunities which are usually considered as important determinants of leverage (DeAngelo and Masulis, 1980). The first equation aims at explaining the level of accounts payable based on the substitution hypothesis. It shows a significantly negative relationship between total loans and accounts payable. This is consistent with the substitution hypothesis which states firms use trade credit, in spite of its supposed high cost, when they are credit constrained. This conjecture is confirmed by the fact that there is a significantly negative relationship between trade credit and the relative amount of unused credit lines, which is a direct indicator of the absence of credit rationing. Moreover, there is a negative relationship between firm size which is an indicator of firm opacity and trade credit. The larger the firm is, the less opaque it is and the less trade credit it uses. As far as the firm s age is concerned, note that both the log of the firm age and the square of the log of the firm age coefficients are significantly different from zero but of opposite sign. Accounts payable first increase with age and eventually fall. More precisely, age is positively related to accounts payable for firms younger than 10 years. For older firms, the effect is reversed. This could be related to the firm s growth cycle. We can suppose indeed that young firms tend to experience a more rapid growth than old ones (Table I shows that young firms total assets are bigger than old ones, which tends to support this conjecture). As a consequence, they should have higher financing needs and be all the more indebted as they do not have substantial retained earnings. On the contrary, old firms may have lower financial needs and may prefer internal financing (resulting from retained earnings) to trade credit, which is consistent with the Pecking Order Theory (Myers and Majluf, 1984). This could also be connected to the fact that trade credit helps young firms to build their reputation as low risk 19

20 borrowers and is used as a high quality signal towards uninformed banks (Biais and Gollier, 1997). Finally, firms use more trade credit when the ownership share of the principal owner is substantial. This could result from the owner s wish to keep control of the firm and use any financial resource even if it is expensive, providing it does not question firm governance structure. As far as the structure of the business is concerned, firm total income and percentage of current assets are positively related to accounts payable. This is consistent with the idea that current assets are financed with current debts. Generally speaking, these results are consistent with those of previous studies (Petersen and Rajan, 1997). They show accounts payable are determined by the structure of the business, possible credit rationing and the firm s reputation. The second equation aims at explaining the level of bank debt based on the complementary hypothesis. It shows that bank debt level grows with accounts payable which suggests that trade credit and bank finance are complementary sources of funds. This could result from the role of trade credit as a signal of a firm s quality, as suggested by Biais and Gollier (1997). Moreover, bank debt is positively related to firm size, which is in accordance with the idea that small and opaque firms have problems raising funds from banks. Our results also show that firms are all the more indebted when they have many depreciable assets. This is consistent with the idea that depreciable assets should be financed with long term funding and this tends to confirm the role of collateral as a means of overcoming information asymmetries (to prevent asset substitution from firms and credit rationing from banks). 20

21 Moreover, firms using capital lease are more leveraged than others. One can then suppose that firms with high financing needs use any available financing resource. Alternatively, the fact that capital lease companies are engaged in the firm could induce banks to lend too. Also, firms located in urban areas seem to be less indebted than firms located in rural ones. As far as firm age is concerned, older firms appear to be less indebted than young ones. This is consistent with the Pecking Order Theory which states that firms should prefer internal financing to debt. Ceteris paribus, old firms shall have accumulated more retained earnings than young ones and rely more on internal financing 22. On the contrary, young firms may have great financial needs due to their high growth and use debt more as they do not have much available internal funding. Some theoretical papers have demonstrated that the banking relationship can improve a firm s access to bank debt (Petersen and Rajan, 1994 ; Cole, 1998 ; Elsas and Krahnen, 1998). Our results show this is the case at the beginning of the banking relationship (when it is shorter than eleven and a half years). Firms can build their reputation as high quality borrowers and alleviate credit constraints. On the contrary, firms seem to be less indebted when they have longer relationships with banks. We do not have any convincing explanation for this reverse effect, except the ones already mentioned to explain the influence of a firm s age. The second equation shows that informational asymmetries seem decisive to explain the level of bank debt. This highlights the role of trade credit as a complementary source of funding which could result from its role as a signal of borrower quality. 22 It is clear that retained earnings can only be accumulated if the firm is profitable. But one can suppose that old firms have been profitable at least at some period of their life. They would not have survived otherwise. 21

22 To summarize, trade credit appears as an alternative source of financing to bank debt when firms are credit rationed. This is consistent with the substitution hypothesis. But bank debt also appears as a complementary source of financing to trade credit. This is consistent with the complementary hypothesis. On the surface, these two hypotheses would seem to be competing, but our results indicate that both hypotheses are consistent with the data. This overall finding would at first blush seem to be impossible. One possibility is that the two hypotheses apply to different classes of small businesses as suggested by our reputationnal hypothesis. In order to test this third hypothesis, we reestimated the model depending on the duration of the longest relationship with a bank. We split our sample to distinguish firms for which there is a positive vs. negative relationship between the relative amount of bank debt and the duration of bank relationship into the second regression. The effect of the banking relationship duration on the amount of bank debt becomes negative after eleven and a half years (compared to ten years for the median value). We chose to split the sample according to this criteria because we consider that it more accurately represents the knowledge the bank should have on the firm during their business relation than others such as firm size or age. Nevertheless, these variables are positively correlated as shown in Table IV. <Insert Table IV about here> We expect the complementary hypothesis to be rejected for firm with a long banking relationship whereas the substitution hypothesis should be validated for all firms. Regressions results are reported in Table V. 22

23 <Insert Table V about here> For firms with a short relationship with a bank, as far as accounts payable are concerned, results are quite similar to those obtained with the sample considered as a whole and consistent with the substitution hypothesis. Considering the determinants of bank debt, results are consistent with the complementary hypothesis. Depreciable assets and accounts payable seem to play a major role in explaining the level of bank debt. We can thus suppose that banks agree to lend money to firms with a short banking relationship when they can offer some collateral and when suppliers are also committed in the firm. This is consistent with the hypothesis that trade credit makes it possible for opaque firms to build their reputation toward banks. This is perfectly consistent with the results of Biais and Gollier (1997) where trade credit is both a consequence of credit rationing and a signal that facilitates access to bank debt. For firms with a long banking relationship, accounts payable seem to be inversely related to debt, which seems to confirm that these firms are credit constrained too and is consistent with the substitution hypothesis. Other determinants of accounts payable are quite similar to those of the sample considered as a whole but they explain a smaller part of the observed variance in the use of trade credit. One can thus suppose that the use of trade credit by firms with a long banking relationship is less determined by informational aspects. Considering bank debt determinants, one can notice that accounts payable coefficient is not significantly different from zero, which lets us suppose that trade credit does not play an important role when firms have a long banking relationship. The complementary hypothesis has to be rejected for these firms. Moreover, depreciable assets seem to play a less important 23

24 role, considering both the obtained coefficient and the tstat. One can suppose indeed that when firms have longstanding relationships with some banks, possible collateral is not such a key factor in obtaining a loan. On the contrary, other variables such as the firm s age or book value of assets, linked to the firm s financial needs, seem to be more important. Globally speaking, informational factors seem to be less important to explain bank debt level. To summarize, one can say that trade credit clearly appears to be a substitute for bank debt, whatever the length of the firm banking relationship. Put another way, banking relationship is not a perfect insurance against credit rationing. Nevertheless, trade credit appears to be a complementary source of funds for firms with a short banking relationship. Hence, we cannot reject the view that the use of trade credit by firms of good quality reflects their will to built their reputation. This is perfectly consistent with the reputationnal hypothesis discussed previously. IV. Discussion and conclusion We now attempt to draw together the evidence obtained concerning the use of trade credit by firms. Firstly, using recent US small business data, we provide new evidence concerning the role of trade credit as a substitute for bank debt. The substitution hypothesis cannot be rejected. Indeed, an increase in bank debt lowers the amount of trade credit the firm uses. This result is robust with any effect induced by the firm s legal form and industry. It can be interpreted as evidence that firms are credit constrained and have to use trade credit when suffering from a restrained access to bank finance. In accordance with Berger and Udell (1998), we find that the difficulties in accessing bank debt are related to firms opacity. This point is confirmed by the fact that credit rationing and information asymmetry proxies do have a positive influence on the use of trade credit by firms. 24

25 Secondly, and new in this paper, we show that trade credit may appear as a complementary source of financing to bank debt. The complementary hypothesis cannot be rejected, except for firms with a long banking relationship. The more trade credit firms use, the more indebted towards banks they are. We interpret this result along the lines of the theory developed by Biais and Gollier (1997) who emphasize the informational role of trade credit. Suppliers have an informational advantage over banks (e.g., Petersen and Rajan, 1997). Suppliers have unique access to the relevant information concerning their customers during the course of their business relation, even if these firms may appear informationally opaque to banks. Some of the suppliers private information is revealed when extending trade credit to business customers, making it possible for banks to update their beliefs concerning the quality of firms. Therefore, banks may lend to firms with positive NPV projects which would not have been funded in the absence of trade credit because of an information gap. This result doesn t hold for firms with a long banking relationship. We suggest that this phenomenon is linked to the fact that, for these firms, the informational gap is at least partially solved thank to the bank informational advantage. Furthermore, this signaling role of trade credit is observed in the US context where the banking system is supposed to be informationally efficient. Therefore, it is not a consequence of the underdevelopment of the financial system, but quite a general phenomenon. However, the analysis in this paper should be considered carefully as it mainly focuses on the signaling role of trade credit. Indeed, several papers have emphasized other important aspects of trade credit that one should take into account in a more general analysis of the use of trade credit by firms à la Petersen and Rajan (1997). For example, more attention should be given to the price discrimination aspect of trade credit or the influence of the trade discount policy (Danielson and Scott, 2000). Our results raise the question of the way trade credit and bank debt can appear both as substitutes and complements for firms that experienced only some short banking relationships. Following the same approach we employed here, one can try to discriminate among these firms or to use some better proxy for the borrower type and reputation. The use of panel data would also allow a more direct test of the reputationnal hypothesis. Besides, tests should be developed in order to discriminate between the different theories that support 25