Title: Trade Credit and Small Distressed Firms

Transcription

1 Title: Trade Credit and Small Distressed Firms Abstract: We analyze trade credit in a sample of small distressed firms that are restructured under the Belgian procedure of court-supervised reorganization (Chapter 11 in the U.S). During the prebankruptcy period, firms rely on trade credit to finance a lack of internally generated cash. This is specifically true if banks contract their lending during the pre-bankruptcy period. Trade creditors lend more if firms are more transparent, and if their board members are less involved with previous bankruptcies. Suppliers may also lend to our sample firms because they have a long-run interest in their survival since our sample firms indeed succeed in getting a plan voted by their creditors and confirmed by the court. Keywords: trade credit; small distressed firms; court-supervised reorganization; bankruptcy JEL: G30, G33, L26 1

2 1. Introduction We analyze the role of trade creditors in small distressed firms that have some prospects of corporate rescue. We specifically test whether suppliers act as liquidity providers in the period before filing a petition for court-supervised reorganization (hereafter the pre-bankruptcy period) in response to a shortfall in internally generated cash and a contraction of the bank debt. We find that entrepreneurs demand more trade credit during the pre-bankruptcy period to finance their loss-making business, and suppliers are willing to provide this credit. As in Peterson and Rajan (1997), the firm s ability to generate cash internally strongly affects the firm s demand for trade credit 1. In line with Peterson and Rajan, it could be argued that cash flow precedes trade credit in the pecking order. Related, Cunat (2007) finds that trade credit is more prevalent when firms experience a liquidity shock that may threaten their survival. In his study, a firm is defined to have a liquidity shock if the total amount of cash and bank deposits scaled by assets drops by more than 10% (that is 10% of the assets of the firm) or if dividends are cut. Our finding that cash flow precedes trade credit in the financing order is robust to controlling for a distressed firm s access to bank debt and government debt 2. Since Franks and Sussman (2005) find that bank debt is substituted by trade credit in the running up to bankruptcy, we specifically analyze the impact of bank behavior on the trade credit levels at the start of the reorganization procedure. We find that if banks contract their lending during the pre-bankruptcy period, distressed firms rely on trade credit to finance a shortfall in cash flow, while this is not the case under a bank debt expansion. This suggests that entrepreneurs use trade credit to fill liquidity gaps created by both a lack of internal profit generation and the contraction of bank debt. This pre-bankruptcy dynamics strongly contributes to the high trade credit percentage of 40.05% observed in our sample firms. Several trade credit theories explain why suppliers may want to extend credit to distressed companies in the first place. Debt enforcement theories suggest that trade creditors may have an advantage in the enforcement of due trade credit. Several papers (see e.g. Santos and Longhofer 2003; Frank and Maksimovic 2004) argue that suppliers access to distribution channels gives them an advantage in liquidating intermediate goods in case of default. Cunat (2007) suggests that suppliers are more able to enforce debt repayment than banks because they can credibly threaten to cut the supply of intermediate goods to their debtors. Equity-stake theories put forward that trade creditors may have a long-term interest in the survival of their customer, which encourages them to contribute to their rescue. Cunat (2007) argues that trade creditors have an equity-like stake in a customer s business as long as the value of their relationship is higher than the cost of helping in the survival of the distressed firm by providing additional trade credit or by extending the maturity of existing trade credit. The trade creditor s implicit equity stake makes them act as liquidity providers, supporting their customers 1 Peterson and Rajan (1997) find that each additional dollar of profits lowers the firm s demand for trade credit by 23 cents. 2 We also control for unpaid taxes and social contributions, which account for a substantial part of the unsecured debt in our Belgian court-supervised reorganization cases. For Chapter-11 firms and especially for small ones, Baird et al. (2007) equally show that unpaid sales taxes and payroll taxes are sizable. They argue that entrepreneurs of small distressed firms have fewer incentives to transfer sales taxes and payroll taxes in the running up to bankruptcy. 2

3 whenever they face temporary liquidity shocks. Wilner s model (2000), where suppliers give larger concessions than banks in the case of debt renegotiation, is also based on the existence of a trade creditor s stake in the survival of their customer. Huyghebaert et al. (2007) investigate the choice between bank debt and trade credit in Belgian start-ups. They suggest that these start-ups rely more on trade credit than bank debt, because banks follow a strict liquidation policy, whereas a supplier s implicit equity stake results in more willingness to renegotiate the outstanding debt or grant additional debt. Information advantage theories suggest that suppliers may have an information advantage over financial institutions in assessing their customers creditworthiness, which allows them to play an active role in financing distressed firms. Specifically trade creditors may get information faster and at lower cost compared to banks because it is obtained in the normal course of business 3. Peterson and Rajan (1997) argue that these debt enforcement theories and the equity-stake theory explain why suppliers are still willing to lend to constrained small firms 4. They find evidence suggesting that firms use more trade credit when credit from financial institutions is limited or unavailable. Franks and Sussman s analysis (2005) reveals some related findings. In a subsample of distressed small to medium sized UK firms that eventually are liquidated, they find that banks rarely expand their credit, while trade creditors are more willing to provide. For every pound sterling that the bank has withdrawn, the trade creditors have put in on average Franks and Sussman rely on the existence of a trade creditors equity-like stake in a distressed company to explain their findings 5. In our sample we can assume that an implicit equity-like exists for all firms, since they all obtain a creditor-confirmed going-concern plan under Belgian court-supervised reorganization (Chapter 11 in the U.S.). We cannot test directly the debt enforcement theory, but we control for its effects by including industry dummies. The information advantage theory is supported since trade creditors lend more if firms are more transparent and if their board members are less involved with previous bankruptcies. This paper is organized as follows. Section 2 discusses the data and gives an overview of the debt composition at the start of the court-supervised procedure. Section 3 formulates the hypothesis and presents the empirical approach, while results are discussed in section 4. Section 5 concludes. 2. Data 2.1. Data sources and sampling procedure Our hand-collected dataset consists of the complete judicial records of distressed firms that eventually obtain confirmed reorganization plans under court-supervised reorganization in Belgium. Approximately 306 plans were confirmed between January 1, 1998 and June 30, 2004 with one of the 3 If trade creditors have private information about a firm s creditworthiness, Biais and Gollier (1997) show how firms can rely on trade credit to convey their private information to the bank, and eventually get additional bank finance. 4 Peterson and Rajan (1997) give a fairly complete overview of all existing trade credit theories. Example given, trade credit can also be used as a tool for price discrimination because it changes the effective price of goods. 5 Franks and Sussman (2005) however argue that trade creditors might be simply unaware of the financial difficulties of their customers, which results in ill-informed suppliers lending to distressed companies. 3

4 23 regional Belgian Bankruptcy Courts. Our sample is restricted to all confirmed reorganization plans submitted to 17 of those Bankruptcy Courts. This amounts to 190 reorganization plans or 62% of the population of confirmed plans. Corporations and sole proprietorships submitted respectively 125 and 65 plans ( = 190). Blocks of closely related corporations jointly submitted five out of those 125 plans 6. The dataset is complemented with pre-bankruptcy financial statement data from the Graydondatabase and the Belfirst DVD s, which are delivered by the private data vendors Graydon Belgium and Bureau van Dijk respectively. We analyze a sample of small distressed corporations that submitted a going concern plan. We exclude corporations with total assets exceeding , which leaves a sample of 107 small corporations. We additionally exclude an incorporated soccer club and one liquidation scheme among the small corporations 7. After these restrictions we retain 105 small corporations in the sample. Our dataset is complemented with financial statement data prior to petition filing for bankruptcy-reorganization. To ensure a sufficiently high quality of the financial statement data, we do not include corporations for which the time period between the financial statement date and the filing date for bankruptcyreorganization is longer then 18 months. This removes another 14 corporations, resulting in a sample of 91 corporations. Since the court jointly appraises the cases of closely related corporations, the data on the financial statements should be aggregated. Simple data aggregation is not recommended though, because of intra-group transactions and consolidated accounts are not available. Plans submitted by closely related corporations are therefore excluded from the sample of corporations resulting in final sample of 89 corporations Sample firms Table 1: Firm characteristics sorted by legal form. N Mean Median Std. Dev. Min Max Public Limited Liability Corporation Pre-bankruptcy total Assets ( 1000) Employees (No.) Liabilities ( 1000) Liabilities/pre-bankruptcy assets Private Limited Companies Pre-bankruptcy total Assets ( 1000) Employees (No.) Liabilities ( 1000) Liabilities/pre-bankruptcy assets The corporations differ by legal form. 45 corporations are non-quoted public limited liability corporations (Société Anonyme), 41 are private limited companies (Société Privée à Responsabilité Limitée), and 3 incorporated firms have another legal status. Table 1 gives summary statistics sorted 6 Five blocks of incorporated firms file jointly a plan. Those blocks respectively consists of 9, 4, 2, 2, and 2 corporations. 139 corporations ( ) are subsequently involved with the 125 plans. 7 Three large corporations confirmed a liquidation scheme, but are already excluded. 8 Three groups were already removed before because total group assets were larger than

5 by legal form. Total liabilities are measured at the initiation of the procedure, i.e. 6 to 9 months before plan confirmation. The public limited liability corporations are clearly larger than the private limited companies. Our sample firms are less underwater compared to those in Bris et al. (2006), likely because we use a sample of confirmed plans like in Baird et al. (2007) The debt structure at the moment of initiation of the procedure Table 2: Debt composition of sample firms as reported in the confirmed reorganization plans. Panel A shows the debt structure of our 89 small corporations at the start of the procedure. In panel B, the sample firms are restricted to 68 firms with (secured and unsecured) bank debt. Panel C provides data on loan securities. These securities provide a contractual liquidation right contingent upon default. A fixed charge is a security in real estate. A floating charge is a security on mainly working capital. Panel A: Debt structure at the initiation of the procedure Mean Median St. Dev. Min Max Number of plans with specific debt Secured bank debt out of 89 Unsecured bank debt out of 89 Trade credit out of 89 Tax & Social Contributions out of 89 Owner-Directors out of 89 Panel B: Debt structure of bank-financed firms at the initiation of the procedure Mean Median St. Dev. Min Max Number of plans with specific debt Secured bank debt out of 68 Unsecured bank debt out of 68 Trade credit out of 68 Tax & Social Contributions out of 68 Owner-Directors out of 68 Panel C: Collateral rights Number of bank-financed firms with... Both a Fixed and floating charge 36 Only a Fixed charge 1 Only a Floating charge 21 No security 10 Total bank-financed firms 68 ( ) Personal guarantee (in addition to other securities) 13 Panel A of table 2 shows that bank debt and trade credit are the main sources of finance in our sample of 89 small corporations and that the latter source dominates the former on average 9. Trade credit amounts on average to 40% of total debt in our sample 10. Only 58 of the 89 distressed sample firms 9 Creditors benefiting from retention of title clauses are most likely trade creditors, and their claims are therefore included in the trade credit. Due employee wages are incorporated in the trade credit because bankruptcy documents do not allow to distinguish them from trade claims. Social security contributions regarding the employee wages are included in the government debt. Clearly, the continuation decision of distressed firms critically depends on the employees, which typically results in paying out wages (but without transferring social contributions to the administration). Fisher & Martel (1994) report that only 23% of Canadian plans involve some wage claims; wage claims to total liabilities amounts to 0.35% in their sample study. 10 Cunat (2007) reports a ratio of trade credit to total debt of 35% for a sample of small US firms (data-input from the NSSBF). For the U.K., he finds a percentage of 41% for a sample of medium and large corporations (data-input from FAME). See Rajan and Zingales (1995) for an analysis of the capital structure for a sample of large listed companies of the G-7 countries. Their accounts payable to assets amounts approx. 15 %. 5

6 have secured bank financing at all at (the moment of) procedure initiation. Due taxes and social claims also constitute a considerable debt mass, while junior-subordinated owner/director debt 11 is not a frequent source of finance. Panel B shows that trade creditors remain the main providers of external funds even for the 68 cases with bank debt (both secured and unsecured). Remarkably, in both panel A and B distressed firms rely heavily on tax and social contributions as a source of finance (more than 20% in both panels). Unpaid government claims are omnipresent and seem larger in Belgium than in other countries like the U.S. or Canada. Bris et al. state that median Chapter 11 tax claims are zero in their sample 12. Their ratio of tax claims on total liabilities depends on the filing district. The ratio averages 14% in New York, while it is only 3% in Arizona. Using a sample of confirmed plans, Baird et al. (2007) report a percentage of 7,3%. Unpaid Canadian government claims average only a few percentages (Fisher & Martel, 1994, 1995). Panel C of table 2 reports that almost 90% of bank debt is covered by a fixed and/or floating charge. These securities provide a contractual liquidation right contingent on default. A fixed charge is a security in real estate. A floating charge is a security on machinery and working capital such as receivables and inventory. This high degree of collateralization is comparable to other European countries (see Davydenko and Franks for the U.K., France and Germany, 2008). In our sample, multiple bank situations occur in only 16 of 89 cases 13, implying that securities are often concentrated in the hands of a single bank. 3. Hypotheses and empirical approach Peterson and Rajan (1997) argue that the demand of trade credit may be driven by arguments from the pecking order theory (Myers 1984). The classical pecking order theory predicts a sequence for financing decisions: firms finance new investments first with internal funds, then with safe debt, then risky debt and finally with outside equity. Adverse selection costs (due to information asymmetries) and transaction costs of issuing risky debt and equity securities induce this hierarchy. Internal funds have no adverse selection problem, while both debt and equity require an adverse selection risk premium. Debt demands a lower risk premium than equity. The nature of the financing sequence of the pecking order theory results in a minimization of the adverse selection premiums and transaction costs 14. Specifically Peterson and Rajan argue that internal funds precede trade credit in the pecking order. They find that each additional dollar of profits lowers the demand for trade credit by 23 cents Owner/director debt includes credit provided by group companies. 12 Tax claims include social contributions in the U.S. Government debt in our study refers to both tax and social contributions. 13 2, 3 and 4 banks are involved with respectively 12, 3 and 1 corporations. 14 The empirical literature on the pecking order theory especially focuses on the financing decisions of public quoted American firms; we refer to Helwege and Liang (1996), Shyam-Sunder and Myers (1999), and Frank and Goyal (2003) in this respect. See Barclay et al. (2006) and Manigart and Van Acker (2009) on capital structure and high growth ventures. 15 Peterson and Rajan (1997) appropriately argue that cash flow rather than profits is the correct variable to test the pecking order theory. 6

7 This result is found in a sample of small firms, of which are 90% owner-managed. Outside equity is however not issued by these small firms, implying that the model of Myers and Majluf (1984) based on informational asymmetries between existing and prospective shareholders is not appropriate to explain the found pecking order sequence. The trade creditor s equity-like stake and other specific trade credit theories (see introduction) may however explain why trade creditors are willing to finance a shortfall in cash flow. We formulate our first hypothesis as follows: Hypothesis 1.: Cash flow precedes trade credit in the pecking order for small distressed firms. The term pecking order in hypothesis 1 refers to the financing sequence implied by specific trade credit theories rather than by the model of Myers and Majluf (1984). In order to test hypothesis 1, we estimate reduced form models in the spirit of Peterson and Rajan (1997) that link the use of trade credit to both demand and supply factors. Specifically we estimate ( TC / TA ) = b + b C + b BD + b GD + b STC + b Z + ε (1) t t 1 i 0 1 i, t 1 2 i, t 1 3 i, t 1 4 i, t 5 i, t 1 i The subscript t-1 refers to the last annual account prior to filing bankruptcy (pre-bankruptcy), t refers to the moment of initiation of the procedure, and the subscript i refers to the firm. The vector C is a vector of proxies for the firm s ability to generate cash through its operations. Peterson and Rajan (1997) find that a firm s ability to generate cash internally (vector C) decreases its demand for trade credit, and this after controlling for the firm s access to credit from financial institutions. We will therefore define various proxies to control for the firm s access to bank debt (vector BD) in the empirical sections. Baird et al (2007) argue that small firms also rely on unpaid taxes to finance their business operations in the running-up to Chapter 11. Since the access to government debt may therefore also affect the demand for trade credit, we include a proxy for unpaid government debt (vector GD) in the analysis. Last we also control for the supply of trade credit (vector STC), by including measures of the quality of the management and the transparency of the firm that are proxies for the information advantage theories of trade credit. Z is a vector of control variables including sector dummies. If b 1 <0, we cannot reject hypothesis 1. Special attention will be devoted to the impact of a contraction or expansion of bank debt during the pre-bankruptcy period on the provision of trade credit. Rodríguez-Rodríguez (2006) suggests that bank debt and trade credit may be substitutes and that firms experiencing short term bank finance constraints may be more likely to use trade credit. If banks contract their lending in the running up to bankruptcy filing, the bank debt capacity of the distressed firm has been reached 16. Specifically under 16 We refer to Lemmon and Zender (2007) for an empirical study of the classical pecking order theory of Myers (1984) after controlling for a firm s debt capacity. They argue that the debt capacity of a distressed firm is reached if the costs of financial distress curtail further debt issues. According to Lemmon and Zender (2007), the combination of debt capacity and the pecking order theory suggests that the costs of adverse selection are dominant for low to moderate leverage levels but that tradeoff-like forces become primary motivators of financing decisions at high levels of leverage. Those so-called tradeoff-like forces refer to the tradeoff theory (that competes with the pecking order theory) to explain the financing decisions of firms in modern corporate finance literature. The tradeoff theory of capital structure predicts that firms choose their mix of debt and equity 7

8 a bank debt contraction, an additional liquidity constraint due to a shortfall in internally generated cash might heavily increase the use of trade credit. In the opposite case of a pre-bankruptcy bank debt expansion, the distressed firm s bank debt capacity is not fully exhausted as banks are still willing to provide additional credit. This bank debt expansion improves the distressed firm s liquidity position, and entrepreneurs are expected to resort less on trade credit to finance a shortfall in internal funds. Hence our second hypothesis: Hypothesis 2: Small distressed firms with low internal cash flows rely more on trade credit it their bank contracts lending in the pre-bankruptcy period. In order to test hypothesis 2, we will estimate a simplified benchmark version of (1) for several subsamples. First we distinguish between a subsample with a bank debt contraction and a bank debt expansion. Later we distinguish between large and small bank debt contractions. We consider two data sources for our bank debt contraction and expansion variables. Hypothesis 2 cannot be rejected if b 1 is larger in the subsample with bank debt contraction than in the subsample with bank debt expansion and is further emphasized if b 1 is larger in the subsample with the largest bank debt contraction 4. Results 4.1. Does cash flow precedes trade credit in the pecking order? In table 3 below, we provide estimates of (1). Trade credit is measured at the initiation of the bankruptcy procedure (data from the judicial files), and is normalized by pre-bankruptcy assets. To verify hypothesis 1, we introduce a number of financial variables. If we find that trade credit is negatively related to the internal money flows, the net demand effect of the pecking order formulated in hypothesis 1 dominates. In specification 1 we introduce profitability (net profits 17 /assets). The negative coefficient confirms more heavily distressed firms demand more trade credit. In specification 2 we substitute cash flow (cash flow/assets) for profitability as a measure of internal money flow generation. Again we find a strongly significant negative relation, supporting the earlier conclusion that more distressed firms demand more trade credit. The estimated cash flow coefficient of implies that each additional euro of loss in cash flow increases the demand for trade credit by around 32 cents (similar for net profits/assets in specification 1). This cash flow variable is introduced in most specifications and the results are very robust. Hypothesis 1 cannot be rejected: trade credit is higher in the pecking order than internally generated cash in our sample of small distressed firms. Aspects of both economic and financial distress drive the variables net profits/assets and cash flow/assets. Earning before interests, taxes, depreciation and amortization (EBITDA) is commonly to balance the benefits and costs of debt. Tax benefits of borrowed money and the control of free cash flow problems are argued to increase the use of debt, while the costs of financial distress and conflicts between debt holders and equity provides firms with incentives to limit their debt financing. A value-maximizing firm equates benefit and costs at the margin. At high levels of leverage, the costs of financial distress typically curtail further debt issues. Lemmon and Zender (2007) argue that the use of debt capacity makes it more difficult to distinguish between the pecking order theory and the tradeoff theory. 17 Net profit before taxes. 8

9 used as a proxy to investigate the degree of economic distress, as it does not reflect differences in debt structure (see Hotchkiss 1995). This operational cash flow variable is scaled by assets and introduced in specification 3. Its estimate is 1.71 standard deviations away from zero and the explanatory power of the model drops to some extent, suggesting that both economic and financial distress, rather than only the former type of distress, drive the demand for trade credit in the pre-bankruptcy period. In specification 4 to 11 we control for the access to bank debt. Peterson & Rajan (1997) include proxies for a firm s credit availability and the relationships with financial institutions in their trade credit models. Firms with large unused lines of credit and with strong bank lending relationships demand less trade credit. The relationship with financial institutions has however no effect on the supply of trade credit. The validity of the pecking order hypothesis might critically depend on the distressed firm s access to bank financing, especially in the bank-based European credit system. The results however show that the coefficient for cash flow in table 3 is consistently negative across specifications 4 to 11. Our finding that b 1 < 0 is therefore very robust to the inclusion of proxies for the access of bank debt. All proxies for the access to bank debt show up with a negative sign, suggesting that bank debt and trade credit are substitutes. Specification 4 controls for the distressed firm s access to bank debt financing by including the variable pre-bankruptcy bank debt (bank debt scaled by assets). Specification 4 also controls for industry effects 18. Cunat (2007) finds that trade credit use is higher in firms with low levels of collateralizable assets as firms with more land and fixed assets have more access to other financing sources. He finds that the economic effect of his collateral variable measured as the book value of land and fixed assets to total assets is particularly strong. In our sample, we find that bank debt is heavily collateralized (see section 3). We show in table 1 of appendix A that the ratio of the book value of land and buildings to total assets is a very good predictor of outstanding bank debt. Specification 5 therefore introduces the variable Land and buildings on total assets as a proxy for a firm s access to bank financing. A significant negative coefficient is found with a relatively large coefficient. The difference between the first and third quartiles of our collateral variable is (0.31 in Cunat), which implies a reduction of trade credit of 10.36% (4.03% in Cunat). Specification 6 additionally controls for industry effects. Specification 7 and 8 are two-stage least squares regressions with the variable Bank debt at procedure initiation/assets as instrumented variable. The instrumental variables are discussed in appendix A, and include the (1) Land and buildings/assets and (2) a proxy for the variability in the business returns 19 (firms with stable cash flows have more access to bank debt) and (3) the dummy variable Debt personally guaranteed. The coefficient of the variable Bank debt at procedure initiation/assets is marginally significant in specification 7, and becomes insignificant after controlling for industry effects in specification 8. One euro of additional bank debt at procedure 18 In the remaining specifications, we will often control for industry effects in main specifications. Industry dummies are defined as follows: wholesale (23 cases), retail (15 cases), manufacturing (13 cases), hotels and restaurants (9 cases), construction (8 cases), and other industries (21 cases). Other industries are the omitted category. We find that manufacturing firms have significantly lower levels of trade credit. 19 We use the industry s variation in profit margin as proxy for the variability in the business returns. The industry s variation in profit margin consists in the industry average of the standard deviation of the operating profit margin over the last 3 fiscal years prior to petition filing. This variable is based on variation in profit margin within businesses over time (i.e. non cross-sectional). 9

10 initiation results in an expected reduction of trade credit use of around 35 and 28 cents in respectively specification 7 and 8. The impact of the access to government debt as informal financing mechanism for distressed firms (see Baird et al 2007) is controlled in specifications 9 to 11. Table 1 of appendix B shows that the variable Payroll costs/assets is a very good predictor of the unpaid government debt level at the moment of initiation of the bankruptcy-reorganization procedure, since overdue taxes and social contributions on salaries are the origin of the unpaid government debt. In specifications 9 to 11 we find a large and negative coefficient for Payroll costs/assets, indicating that trade credit and overdue government debt are substitutes. After controlling for a distressed firm s access to both government debt and bank debt, we find that each additional euro of loss in cash flows increases the demand for trade credit by even more than 30 cents. Trade creditors do not supply trade credit blindly to any firm that demands it. In all specifications we include the log of firm age and a dummy for Public Limited liability Corporation. Both variables are a proxy of the transparency of the firms. We indeed find that Public limited Liability Corporations get more trade credit from their suppliers, but older firms do not. In specifications 10 to 11 we enrich the supplier s information set with information about the management of the distressed firm. We tracked down the involvement of members of the executive board in earlier bankruptcies in Belgium (Previous bankruptcies) and the experience of the executive board of the distressed firm on the boards of other Belgian firms (Management experience). The former variable counts earlier bankruptcies in which the board of directors has been involved as a director 20, while the latter counts all positions on boards ever held by members of the distressed firm s board, excluding earlier bankruptcies. We find that suppliers extend less trade credit to firms whose managers have more bankruptcies on their slate. Suppliers may also tend to extend more trade credit to firms with a more experienced management team, although this result is not significant. All specifications include a control for firm age. Larger firms have less trade credit in all specifications. It could be that larger firms demand less trade credit because they have more access to other financing sources (due to limited information asymmetries) or because they have less growth opportunities, as suggested by Peterson & Rajan (1997). Hart & Moore (1994) and Diamond (1991) present rationales for firms matching the maturity of assets and liabilities. Indeed, our control variable for current assets (current assets excluding cash/assets) is clearly related to trade credit levels, although its significance is specification sensitive. In specifications 5 to 11, the variable Current assets excl. cash/assets is replaced by two main components of current assets: Inventories/assets and Receivables/assets. In this way we avoid the simultaneous introduction of variables that are highly correlated (Current assets excl. cash/assets and Land and buildings/assets). The variables Inventories/assets and Receivables/assets are not significant and the latter has even a negative sign If a firm goes bankrupt two years after management dismissal, we consider the dismissed manager responsible and count it as an involvement in a previous bankruptcy. In Belgium, from a legal point of view, replaced managers even remain responsible for three years after their discharge. 21 This suggests that the positive coefficient on current assets excl. cash / assets found in specifications 1 to 4 is largely driven by the collateral variable (likely idem in the study of Peterson and Rajan). 10

11 Table 3: Determinants of Trade credit/total assets. The dependent variable Trade Credit/Total Assets is the firm s trade credit at the start of the procedure scaled by pre-bankruptcy assets. Apart from the variables Bank debt at procedure initiation/assets, L(Firm Age), D-Public Limited Liability Corporation, Previous Bankruptcies, Management Experience, all variables are directly obtained from the latest annual account prior to the filing for bankruptcy-reorganization. The numerator of the variable Bank debt at procedure initiation / assets is obtained from the judicial records (i.e. at the moment of procedure initiation). The variable Land and buildings/assets is used as collateral proxy (like in Cunat 2007) and determines the firm s access to external bank financing. The variable Bank debt at procedure initiation/assets is instrumented using a simplified bank debt model described in appendix A (specification 6 of table 1 of appendix A). The variable Payroll costs/assets is used as proxy for unpaid taxes and social contributions (see appendix B for a simple model of government debt). The variables Previous Bankruptcies and Management Experience respectively amount to the number of earlier bankruptcies (of other Belgian firms) in which the board of directors has been involved, and their number of past and current management positions in the board of other Belgian firms. The values in brackets are the robust t-statistics: * / ** / *** significant at 10% / 5% / 1%. Spec. 1 Spec. 2 Spec. 3 Spec. 4 Spec. 5 Spec. 6 Spec. 7 Spec. 8 Spec. 9 Spec. 10 Spec. 11 Internal money flow generation Net profits / assets [-2.16]** Cash flow / assets [-2.63]*** [-2.24]** EBITDA / assets [-1.71]* Access to external funds Bank debt / total assets [0.10] Land and buildings / total assets (collateral) [-2.61]** [-2.19]** [-3.04]*** [-2.20]** [-1.57] Bank debt at procedure initiation / assets [-1.66]* (instrum.) [-2.53]** [-3.59]*** [-2.43]** [-1.02] (instrum.) Payroll costs / assets Supply of trade credit D-Public Limited Liability Corporation [1.44] L(Firm Age) [-0.52] [-2.39]** [-3.82]*** [-2.52]* [-2.97]*** [1.40] [1.76]* [1.87]* [1.56] [2.05]** [1.56] [2.19]** [1.47] [1.13] [-0.89] [-0.29] [-1.30] [-0.34] [-0.85] [-0.98] [-1.25] [-0.56] [0.00] Previous Bankruptcies [-2.80]*** Management Experience [1.23] [-3.82]*** [-2.52]* [-2.97]*** [1.13] [0.00] [-2.80]*** [1.23] 11

12 Continuation table 3 Spec. 1 Spec. 2 Spec. 3 Spec. 4 Spec. 5 Spec. 6 Spec. 7 Spec. 8 Spec. 9 Spec. 10 Spec. 11 Controls L(Total Assets) [-3.21]*** [-3.28]*** [-2.90]*** [-2.74]*** [-2.64]*** [-2.51]** [-2.80]*** [-2.93]*** [-2.45]** [-2.90]*** [-2.90]*** Current assets excl. cash/ assets [2.24]** [1.66] [1.98]* [1.38] Inventories / assets [0.42] Receivables / assets [-0.89] [0.19] [-0.91] [0.75] [-0.74] [0.42] [-0.91] [0.44] [-0.38] [0.28] [-0.39] [0.28] [-0.39] Industry dummies NO NO NO YES NO YES NO YES NO NO YES Intercept [4.59]*** [5.08]*** [4.44]*** [5.55]*** [6.24]*** [6.67]*** [6.57]*** [6.54]*** [6.43]*** [6.79]*** [6.79]*** No. of observations R² (Centered R²) (Centered R²)

13 4.2. The effect of pre-bankruptcy bank behavior Hypothesis 2 states that distressed firms rely more on trade credit to finance a shortfall in cash flow if bank debt contracts than if bank debt expands. We test this by estimating a simplified benchmark specification of (1) for various subsamples of pre-bankruptcy bank debt contraction and expansion. The sample split estimations presented in table 4 are based on the difference between the average pre-bankruptcy bank debt, calculated using the annual accounts three years prior to filing, and the bank debt reported in the confirmed plan. Banks reduced their credit in 47 cases, while additional credit was provided in 36 cases 22. In panel A of table 4 we report estimates for the subsample with pre-bankruptcy bank debt contraction (left three columns) and prebankruptcy bank debt expansion (right three columns). We find that trade credit is negatively and significantly related to the internal money flow generation for the 47-case subsample with bank debt contraction, where there is no such relation in the subsample with bank debt expansion. Panel B distinguishes between large and small bank debt contractions, using the median contraction during the 3-year pre-bankruptcy period as cut-off point. The size of the contractions is calculated as the difference between the volume of debt reported in the reorganization plan and the average pre-bankruptcy bank debt level reported in the accounts, scaled by the latter bank debt level. The median contraction amounts %. Under large bank debt contractions entrepreneurs rely more heavily on trade credit to finance a shortfall in cash, while the results do not point in this direction under small bank debt contractions. Hypothesis 2 cannot be rejected. The sample split estimations presented in table 5 rely on unique data on the bank debt flows during a 12-month pre-bankruptcy period, obtained by intermediation of the Central Corporate Credit Register of the National Bank of Belgium. These data were only available for 51 firms (out of 61 cases with bank debt) 23. Banks contract their lending during a 12-month pre-bankruptcy period in 37 cases, while they expand it in 14 cases. In panel A of table 5 we report estimates for the subsample with pre-bankruptcy bank debt contraction (left three columns) and pre-bankruptcy bank debt expansion (right three columns). Trade credit is negatively related to the internal money flow variables in the subsample of debt contraction, though not always significantly, while there is again no negative relation in the subsample with bank debt expansions. In panel B, the median pre-bankruptcy contraction 24 (-22,8%) is employed as cut-off point to define a subsample with large contractions (19 out of 37) and one with small contractions (18 out of 37). Again results indicate that large bank debt contractions push entrepreneurs to heavily rely on trade credit to finance a shortfall in cash, while there is no such evidence if the bank debt contraction is small. Again hypothesis 2 cannot be rejected. 22 There are 6 cases without a change in the bank debt during the pre-bankruptcy period. 23 Data are obtained by intermediation of the Central Corporate Credit Register of the National Bank of Belgium. Missing cases are largely due to small credits (< ), which are not reported in the register. 24 The flows are calculated as the difference between the bank debt at procedure initiation and the bank debt 12 months before filing for bankruptcy-reorganization, scaled by this latter bank debt level. 13

14 Table 4: The effect of pre-bankruptcy bank behavior on the levels of trade credit (sample split 1). The dependent variable Trade Credit/Total Assets is the firm s trade credit at the start of the procedure scaled by pre-bankruptcy assets. To define the subsamples in panel A, we rely on a comparison between the average pre-bankruptcy bank debt, as reported in the accounting data three years prior to petition filing, and the bank debt reported in the reorganization plans. Banks reduced their credit for 47 cases, while additional credit was provided for 36 cases. Panel B defines subsamples with large and small bank debt contractions during the 3-year pre-bankruptcy period. The median contraction is used as cut-off point to define a subsample with large contractions (24 cases) and one with small contractions (23 cases). The values in brackets are the robust t-statistics:* / ** / *** significant at 10% / 5% / 1%. Panel A Bank debt contraction Bank debt expansion Internal funds EBITDA / assets Spec. 1 Spec. 2 Spec. 3 Spec. 1 Spec. 2 Spec. 3 [-2.49]** Net profits / assets [-2.00]** Cash flow / assets Controls L(total assets) Land and buildings / assets (bank debt collateral) [-2.23]** [-2.18]** [-2.11]** [-2.04]** [-1.71]* [-2.03]** [-2.21]** [0.05] [-2.20]** [-2.33]** [-0.15] [-2.13]** [-2.34]** [-0.14] [-2.18]** [-2.32]** Intercept [4.20]*** [3.92]*** [4.04]*** [4.79]*** [4.43]*** [4.62]*** Number of obs R-squared Prob > F *** *** *** *** *** *** Panel B Large bank debt contraction Small bank debt contraciton Spec. 1 Spec. 2 Spec. 3 Spec. 1 Spec. 2 Spec. 3 Internal funds EBITDA / assets [-2.81]** [-0.04] Net profits / assets [-2.40]** [0.02] Cash flow / assets [-1.92]* [0.01] Controls L(total assets) [-1.17] [-1.28] [-1.18]** [-1.37] [-1.55] [-1.51] Land and buildings / assets (bank debt collateral) [0.29] [0.02] [0.19] [-1.42] [-1.49] [-1.57] Intercept [2.38]** [2.31]** [2.39]** [2.17]** [2.23]** [2.27]** Number of obs R-squared Prob > F ** *

15 Table 5: The effect of pre-bankruptcy bank behavior on the levels of trade credit (sample split 2). The dependent variable Trade Credit/Total Assets is the firm s trade credit at the start of the procedure scaled by pre-bankruptcy assets. To define the subsamples in panel A, we rely on unique data on the bank debt flows during a 12-month pre-bankruptcy period, which are obtained by intermediation of the National Bank of Belgium, and were available for 51 firms (out of 68 cases with bank debt). Banks reduced their credit for 37 cases, while additional credit was provided for 14 cases. Panel B defines subsamples with large and small bank debt contractions during the 12-month pre-bankruptcy period. The median contraction is used as cut-off point to define a subsample with large contractions (19 cases) and one with small contractions (18 cases). The values in brackets are the robust t-statistics:* / ** / *** significant at 10% / 5% / 1%. Panel A Bank debt contraction Bank debt expansion Internal funds EBITDA / assets Spec. 1 Spec. 2 Spec. 3 Spec. 1 Spec. 2 Spec. 3 [-1.63] Net profits / assets [-1.52] Cash flow / assets Controls L(total assets) Land and buildings / assets (bank debt collateral) [-3.78]*** [-1.24] [-3.75]*** [-1.37] [-2.12]** [-3.97]*** [-1.31] [0.91] [-0.86] [-1.23] [0.71] [-0.74] [-1.20] [0.84] [-0.83] [-1.22] Intercept [5.42]*** [5.58]*** [5.78]*** [1.49] [1.40]*** [1.44] Number of obs R-squared Prob > F *** *** *** Panel B Large bank debt contraction Small bank debt contraction Spec. 1 Spec. 2 Spec. 3 Spec. 1 Spec. 2 Spec. 3 Internal funds EBITDA / assets [-3.93]*** [0.39] Net profits / assets [-1.98]* [-0.36] Cash flow / assets [-3.71]*** [-0.16] Controls L(total assets) [-3.55]*** [-1.88]* [-3.02]*** [-2.43]** [-2.35]** [-2.37]** Land and buildings / assets (bank debt collateral) [0.19] [-0.45] [-0.15] [-1.51] [-1.52] [-1.54] Intercept [5.78]*** [3.36]*** [5.34]*** [3.28]*** [3.11]*** [3.16]*** Number of obs R-squared Prob > F *** *** ** * * 15

16 5. Conclusions Small distressed firms demand more trade credit during the pre-bankruptcy period to compensate a shortfall in internally generated cash. This is especially the case if banks contract their lending during the pre-bankruptcy period. Bank debt and government debt are found to be substitutes for trace credit: firms with higher bank debt and government debt capacity demand less trade credit. Trade creditors may be willing to play an active role in financing distressed firms because they are well-informed. We indeed find that trade creditors do not supply credit blindly to distressed firms demanding it. Trade creditors are more eager to supply trace credit to Public Limited Liability Corporations which are likely to be more transparent. They are also less eager to supply trade credit to entrepreneurs that were involved in earlier bankruptcies (in other boards than the distressed sample firms ), suggesting that the reputation of the debtor does play an important role. Suppliers might also want to extend trade credit to our sample firms, because all sample firms had their going-concern plan confirmed under court-supervised reorganization. For these plans, the business activity is expected to continue, and the going concern value can be split between the entrepreneur and the creditors. Cunat (2007) shows that if there is a divisible surplus between the suppliers and the entrepreneur, the trade creditors may act as liquidity providers, insuring against liquidity shocks that could endanger the survival of their customer relationships. 16

17 Appendix A. Bank debt financing: a simple model. Financial contracts critically depend on the liquidation value of the pledged assets (see e.g. Hart & Moore 1994, 1998; Berglöf & Von Thadden 1994). The creditor s willingness to provide credit increases with the pool of collateralizable assets. Less-specialized assets are preferred as collateral to avoid fire sales because of illiquid markets upon liquidation (Shleifer & Vichny, 2002). Berger et al. (1996) find that less specialized assets results in more liquidation option value per dollar of book value. Land and buildings are typically considered as very redeployable assets because of their less-specialized nature 25. We expect that more bank debt is attracted when firms have more land and buildings on their balance sheets. In specification 1 of table 1 of this appendix, we regress the book value of land and buildings on the bank s loan size (both variables scaled by pre-bankruptcy assets). The book value of real estate is reported in the pre-bankruptcy accounts and the outstanding bank debt is measured at the moment of imitation of the bankruptcy-reorganization procedure. We control for pre-bankruptcy assets. The empirical findings of specification 1 are in line with our expectations. Specification 2 shows that Machinery and equipment/assets has no effect on the outstanding bank debt. Creditors are reluctant to provide credit when a distressed firm s profit and cash flow realization is highly uncertain (see e.g. Lemmon & Zender 2007). After all, creditors bear the full risk if the distressed business ultimately fails, and they need to share the potential business surplus value with the shareholders. Specification 3 controls for uncertainty by introducing the variation in the industry s profit margin 26 and the industry attrition rate. The latter rate is the proportion of small businesses within a particular industry that file a petition for bankruptcy-liquidation each year (see Morrison 2007). We also introduce EBITDA/assets in specification 3. We find that firms with more variation in the industry s profit margin significantly attract less bank debt. Asymmetric information reduces the willingness of creditors to provide credit. Firm age is used as a measure of the informational transparency of a firm, whereby older firms are expected to be more transparent. Specification 4 introduces the logarithmic value of the age of a firm, i.e. the variable L(firm age). We surprisingly find that older firms have lower levels of bank debt. Specification 5 shows that firms older than 20 years have less bank financing, while young firms (< 5 years) do not suffer from informational asymmetries. 25 Ronen & Sorter (1972) classify land and buildings as less specialized than other fixed assets. 26 The industry s variation in profit margin consists in the industry average of the standard deviation of the operating profit margin over the last 3 fiscal years. This variable is based on variation in profit margin within businesses over time (i.e. non cross-sectional based on 3-digit Nace codes). 17