WPS5328 Policy Research Working Paper 5328 Trade Credit Contracts Leora Klapper Luc Laeven Raghuram Rajan The World Bank Development Research Group Finance and Private Sector Development Team June 2010 Policy Research Working Paper 5328 Abstract This paper provides new evidence on the unique role assuring buyers of product quality; and 4) as a screening of trade credit and contracting terms as a way for both mechanism to gauge buyer default risk. In particular, sellers and buyers to mange business risk. The authors the analysis finds that the largest and most creditworthy use a novel and unique dataset on almost 30,000 supplier buyers receive contracts with the longest maturities, as contracts for 56 large buyers and more than 24,000 measured by net days, from smaller, investment grade suppliers in Europe and North America. The sample suppliers. In comparison, early payment discounts of buyers and suppliers includes firms of varying size, seem to be used as a risk management tool to limit the investment grade, and sectors. The paper finds evidence potential nonpayment risk of trade credit. Early payment in support of four important, and not mutually exclusive, discounts are generally offered to smaller, non-investment reasons for trade credit: 1) as a method of financing; grade buyers. The results suggest that contract terms are 2) as a means of price discrimination; 3) as a bond jointly determined by supplier and buyer characteristics. This paper--a product of the Finance and Private Sector Development Team, Development Research Group--is part of a larger effort in the department to study Capital Structure and Supply Chain Finance. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at lklapper@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Trade Credit Contracts Leora Klapper, Luc Laeven, and Raghuram Rajan * JEL Classification: G32 Keywords: Trade Credit; Capital Structure; Contract Theory; Risk Management * Klapper is Senior Economist at the World Bank. Laeven is Deputy Division Chief at the International Monetary Fund and Research Fellow at CEPR. Rajan is the Eric J. Gleacher Distinguished Service Professor of Finance at the University of Chicago and Research Associate at NBER. We thank PrimeRevenue for generously sharing this data, John Sculley and Shane Maine for helpful comments, and Teresa Molina and Douglas Randall for excellent research assistance. This paper's findings, interpretations, and conclusions are entirely those of the authors and do not necessarily represent the views of the World Bank, the IMF, their Executive Directors, or the countries they represent. I. Introduction Trade credit is an important source of external financing for both small and large firms around the world (Demirguc-Kunt and Maksimovic 2005). For instance, firms often offer working capital financing to their customers, reported as accounts receivables, even if they are small or credit constrained (McMillan and Woodruff 1999, Marotta 2005, and Van Horen 2005). Trade credit might also be available when other forms of financing are not. For instance, during periods of contractions in bank credit (say, due to monetary tightening or bank distress), buyers might depend more on trade credit for short-term financing as a response to credit market tightening, and this may be especially true for small firms (see, for example, Himmelberg et al. 1995, Choi and Kim, 2005, and Love et al. 2007). Trade credit is also thought of as a way for a supplier to engage in price discrimination, giving favored or more powerful clients longer terms (see, for example, Wilner 2000, Fisman and Raturi 2004, Van Horen 2005, and Burkart, Ellingsen, and Giannetti, 2008). Furthermore, trade credit may simply be customary in an industry, with customs driven by economic rationales such as allowing buyers time to check out the quality of the supplied goods (Lee and Stowe, 1993). Studies have explored the supply and demand of trade credit around the world (for instance, Petersen and Rajan 1997, Johnson, McMillan and Woodruff 2002, Boissay and Gropp 2007, and Fabbri and Klapper 2008). Yet, even the stylized facts on how the contract terms of trade credit vary across buyers and suppliers of different characteristics are poorly understood, in part because firm-level contract-level data have not been easily available. For example, what is the typical duration for which trade credit is offered? Who is offered a longer time to pay? Which firms are offered early payment discounts? By whom? This paper addresses these questions using a unique database that includes detailed contract information for about 30,000 transactions for 53 large buyers in the United States and Europe. We summarize typical trade credit terms and how they relate to buyer and seller characteristics. We also study the use of early payment discounts. We then draw some 2 conclusions on the robustness of theories that might have some merit in explaining the patterns in the data. Empirical work on trade credit thus far has been hampered due to a lack of firm-level data on trade credit contract terms. Most studies have used the Federal Reserve Survey of Small Business Financing (SSBF) database of U.S. firms, which has only limited data on credit terms and firm characteristics (e.g., Petersen and Rajan, 1997, and Burkart et al., 2008). An exception is Ng et al. (1999) who use survey level data on 950 listed U.S. firms to study the determinants and characteristics of trade credit contracts. They document wide variation across industries in credit terms, and find that variables capturing buyer and seller reputation are significant determinants of a firm's choice to extend credit, consistent with theories that explain credit terms as contractual solutions to information problems concerning product quality and buyer creditworthiness. Our paper differs from the work by Ng et al. (1999) in several important ways. First, we use unique data on actual trade credit contracts rather than survey based data. Second, we cover a broader set of industries that includes technology firms. Third, our dataset includes trade credit terms not only for US firms but also for international firms. Fourth, we have bilateral contract information, so we can control both for buyer and supplier firm characteristics. Fifth, and importantly, we have multiple contracts for the same firm, rather than a firm-average response, allowing us to include firm fixed effects in our empirical analysis, thereby abstracting from time- invariant firm characteristics that determine the choice of credit terms. Finally, another important distinction is that the Ng et al. data covers only large, publicly traded firms while our dataset includes suppliers of all size. This is an important difference because the credit terms offered by large firms are likely to be very different than those by small firms, especially when credit is extended by small suppliers to large buyers. In short, the bilateral, multi-contract nature of our dataset is truly unique and is a major improvement on datasets that have previously been used to study the determinants of credit terms used in trade credit. This multi-contract structure of our dataset allows us to abstract from unobserved buyer and supplier firm characteristics, something previous empirical work has not been able to do. 3 Using this unique dataset, we find that the largest and most creditworthy buyers receive contracts with the longest maturities, as measured by net days, from smaller, investment grade suppliers. In comparison, early payment discounts seem to be used as a risk management tool to limit the potential nonpayment risk of trade credit. In particular, early payment discounts are generally offered to smaller and non-investment grade buyers, where nonpayment risk tends to be greater. Our results suggest that contract terms are jointly determined by supplier and buyer characteristics. The paper continues as follows. Section II presents summary statistics of our unique data on trade credit contracts. Section III reviews theories of trade credit. Section IV presents the empirical results. Section V concludes. II. Data and Summary Statistics We use a novel database of trade credit contracts for close to the universe of suppliers of 56 large buyers. 1 The data are provided by PrimeRevenue, an online network that links large, global companies, their suppliers, and third-party financial institutions, via the Internet. PrimeRevenue provides software and an IT platform for buyers to post their invoices directly; and suppliers to choose whether to be paid at maturity of the contract or to factor the contract and be paid immediately at a discount. Our data are a snapshot of outstanding receivables as of December 1, 2005. Importantly, this snapshot is before PrimeRevenue started factoring the receivables. Buyers generally post invoices for all `important' suppliers, which is estimated by PrimeRevenue to capture over 90% of total inputs to the buyer. Our database includes information for 29,019 contracts, which includes 56 large buyers and 24,140 suppliers. This includes multiple supplier contracts within and across buyers. The data include complete information on contract terms: spend (contract amount), net days (days within which the buyer has to pay the amount owed), discount days (days within 1 Because purchasing history is proprietary information, we do not know the identity of buyers in our sample. However, as discussed in this section, PrimeRevenue provided us with buyer characteristics (such as size, sector, and location) and the fact that almost all buyers in our sample are Fortune 500 . 4 which the buyer has to pay to get the full discount), discount rate (the size of the discount if the amount is paid by the discount date), and currency. For buyers, we can control for asset size (buckets)2, location (North America or Europe), sector, and whether the buyer is investment grade. For suppliers, we know the asset size (bucket) and whether the supplier is investment grade. Information is not provided on the suppliers' location or sector, since most contracts are written to a local distributor and/or paid to a local bank account. We impute additional variables such as the total buyer spend (as measured by the total dollar amount of contracts entered into) and number of suppliers, and the ratio of spend on a particular supplier to total buyer spend as indicators of the overall importance of the supplier to the buyer's supply chain. Apart from missing information about net days for 832 out of 29,019 contracts, we have complete information on contract terms. We begin by summarizing the main characteristics of buyers, suppliers, and contracts. Table 1 shows summary statistics of buyer and supplier characteristics. First, the buyers in our sample are very large ­ we find that 84% of buyers (weighted by number of contracts) have over US$ 10 billion in sales and less than 1% of buyers have less than US$ 2 billion in sales. The buyers are also creditworthy as measured by whether or not they are investment grade ­ about 84% of buyers in the dataset are investment grade. Buyers are active in a range of industries, with the majority in retail industries. The sectoral distribution in terms of number of contracts is: 6% in auto manufacturing, 34% in diversified retail, 13% in diversified manufacturing, 6% in retail groceries, 11% in retail hard goods, 8% in retail soft goods, 18% in technology, 2% in food and beverages, and 2% in the utility sector. The data encompasses only one firm in the utility sector and two firms in the food and beverages sector. Approximately 59% of buyers (weighted by the number of contracts) are from North America (the US or Canada) and 41% of buyers are from Europe. 2 Buyer and supplier size buckets are (in US$): less than $.1 billion; $0.1-2 billion; $2-7 billion; $7-10 billion; larger than $7 billion. 5 In comparison to the buyers, our suppliers are relatively small: Almost half the suppliers (weighted by the number of contracts) have less than US$ 100 million in sales and only 20% of suppliers have more than US$ 2 billion in sales. Creditworthiness is also an issue for many suppliers, given that almost two-thirds of suppliers are not investment grade. Table 2, Panel A shows summary statistics of contract characteristics. We have a wide distribution of contract amounts varying from about US$ 400 dollars to over US$ 6.5 billion dollars. Contracts in our sample are generally very long in duration ­ the average and median is 60 net days. About 75% of contracts in our sample have net days longer than 30 days, which is longer than the `typical' contract of 30 days previously shown in the literature (Ng, Smith and Smith 1999). For example, 20% of contracts have net days of exactly 30 days, 28% have net days of exactly 60 days, and 17% have net days of exactly 75 days. About 60% of contracts in our sample are denominated in US dollars, followed by almost 40% in euros; this is approximately consistent with the distribution of buyers in Europe and North America (41% and 59%, respectively, as shown in Table 1). In our sample, 13% of contracts (or 3,717 in total) offer early payment discounts.3 Panel B of Table 2 shows summary statistics for this subsample of contracts. We also examine the discount terms, including discount days and discount rate. Almost two-thirds of discount terms are 30 days or less, while 27% are between 30 and 60 days, and 9% are more than 60 days. For example, 20% of discount days are 10 days, 20% are 30 days, and 16% are 60 days. The majority of contracts have a spread of net days less discount days equal to 20 or 30 days (15% and 27%, respectively). The mean ratio of discount to net days is 63%. There is a surprising relationship between net days and discount days ­ over 30% of contracts have a spread of exactly one day. This might suggest that discounts can be used simply to encourage prompt payments, or as an implicit price discount, i.e. an alternative to a cut in list prices.4 The mean and median discount is equal to 2%, and 36% of contracts with a discount have a discount rate equal to 1% or less; and only 8% of contracts with a discount have a discount greater than 2%. 3 This is a comparable figure to that obtained using SSBF survey data on U.S. firms, indicating that 20% of firms that use trade credit are offered an early payment discount from their suppliers. 4 Anecdotally, large buyers do not pay late fees to their suppliers. 6 Trade credit appears expensive for most buyers. The mean effective interest rate, defined as the implied interest rate if the buyer does not pay on the discount date, foregoes the discount, and pays on the due date, is 1 (1 discount rate ) 360 /( net days discount days ) 1 . It is high at 53%, though it varies from a low of 2% to a high of 100%. Table 3 shows the distribution of contract terms by buyer and supplier characteristics. Larger buyers tend to make purchases with a wider range of spend, including more frequent relatively small purchases of less than US$ 1 million in size. Across industries, auto manufacturing and retail hard goods have relatively larger average spend, especially relative to technology, where almost 75% of contracts are less than US$ 1 million in size. We find no notable differences in spend size across buyer location or investment grade. In addition, large suppliers appear to make large sales (and vice versa), while whether a supplier is investment grade does not seem related to average spend. We also examine the relationship between contract maturity and buyer and seller characteristics. We find that the number of net days offered is almost twice as long if the buyer is investment grade, although supplier creditworthiness seems to have no effect on the number of net days offered. We find strong sectoral effects: 85% of contracts in retailing of soft goods have a maturity of 30 days or less, relative to other sectors with longer average maturities. Contracts to firms in Europe are on average longer than contracts in North America (although the sectoral distribution is relatively even across regions). Finally, contracts to the largest and most creditworthy buyers receive longer maturities. Next, we focus on the decision to extend early payment discounts (Table 4). Overall, 13% of buyers (and 13% of contracts) are offered early payment discounts. In general, the buyers receiving a discount are small and non-investment grade, while suppliers offering a discount tend to be larger and are roughly equally likely to be investment or non-investment grade. Suppliers are also most likely to offer discounts to buyers that retail in hard goods. In addition, eight buyers are never offered discounts, while 21 buyers are always offered discounts. In the empirical analysis of this paper we therefore also check how the results look if we drop the firms who never report discounts. 7 Discounts do not appear strongly related to buyer or supplier characteristics, with the exception that very high discounts are more common in the auto industry and among grocery firms. Discount days, the number of days the buyer has to pay and receive a discount, appears strongly related to buyer size ­ 78% of firms with less than US$ 10 billion in total sales have discount days of 30 or less, while only about 64% of firms larger than US$ 10 billion in size receive a short discount window. The mean of net days is 60 days for contracts without discounts and 44 days for contracts with discounts, suggesting that suppliers trade discounts for net days. Importantly, our database also allows for both supplier and buyer fixed effects. About 25% of suppliers (or 7,273 suppliers) sell to multiple buyers. Of these, 3,126 suppliers sell to 2 buyers and 4,147 suppliers sell to 3 or more buyers. In addition, 16% of suppliers (or 4,557 suppliers) have more than one contract with the same buyer. Specifically, 2,685 suppliers have exactly 2 contracts with a buyer, and 1,872 suppliers have 3 or more contracts with a buyer. In general, we find variation in net days and the decision to extend an early payment discount across contracts of a single supplier. III. Theories of Trade Credit Before we explore the data econometrically, it might be useful to outline various theories of trade credit and formulate some testable hypotheses. Much of the work on trade credit has seen it as a form of financing that can overcome traditional impediments in financing. In particular, the seller may know more about, and have more clout over, the buyer than other arm's length financiers (see, for example, Smith, 1987, Brennan et al., 1988, Petersen and Rajan, 1997, Biais and Gollier, 1997, and Burkart and Ellingsen, 2004). Therefore trade credit may be available when other forms of financing are not. Much of this literature argues that large suppliers have a comparative advantage in obtaining outside finance and pass on this advantage to small, credit constrained firms (e.g., Boissay and Gropp, 2007). Similarly, large suppliers may act also as liquidity providers, insuring against liquidity shocks that could endanger the survival of their customer relationships with smaller firms (see, for example, Cunat, 2006). 8 Nevertheless, it is clear from previous studies and our own, that trade credit is not only used to finance credit constrained firms.5 For instance, large, listed, multinational firms around the world, which are unlikely to face financing constraints in the market, hold large volumes of accounts payable on their balance sheet (e.g., Demirguc-Kunt and Maksimovic, 2005). Globally, it is estimated that trade credit financed 90% of world merchandise trade in 2007, valued at about US$ 25 trillion dollars.6 Why might large, investment grade buyers choose to use trade credit financing? One answer is that their suppliers may have cheaper access to financing, and a comparative advantage in passing it on (see Ng et al., 1999). However, as shown in our data set, most suppliers are much smaller than their buyers, and are unlikely to have access to cheaper financing. Another possible explanation is that large buyers receive very favorable contract terms, which reduce their overall borrowing costs. Why small suppliers may want to borrow at high cost in order to provide such cheap financing seems less clear ­ could they not simply offer more of a price discount up front, without incurring the deadweight costs of intermediation? One explanation may be that a country's laws may not allow a vendor to offer different prices to different clients. 7 To the extent that price discrimination is prohibited, variations in trade credit terms also offer opportunities for sellers to offer better terms to more important suppliers (e.g., Brennan et al. 1988). This is consistent with the literature that large buyers can use their market power to demand favorable trade credit terms from their suppliers (see, for example, Burkart, Ellingsen, and Giannetti, 2008, and Fabbri and Klapper 2008). Trade credit might also be used as a risk management mechanism to reduce informational asymmetries between buyers and sellers. Such risk management can either serve to assure product quality or to gauge buyer default risk. 5 In fact, Schiff and Lieber (1974) argue that risk management and inventory management decisions are often taken separately from financing decisions and by different units of the firm, and that consequently trade credit cannot be solely explained on financing grounds. 6 "World Bank urged to lift trade credit finance," Financial Times, November 11, 2008. 7 For example, the Clayton Act in the US prohibits price discrimination across customers for the same good. 9 In terms of reducing uncertainty concerning product quality prior to payment, buyers might demand a credit period. This may be particularly relevant in cross-border sales between different jurisdictions. In this case, trade credit does not play a financing role but can be seen as a warranty that guarantees product quality (see, for example, Lee and Stowe, 1993, and Long et al., 1993). The credit period offers the buyer time to test the quality of the product before deciding whether or not to make payment and accept the merchandise. Alternatively, buyers that take cash discounts and pay early effectively bear product risk. Trade credit can also be offered by suppliers as a screening mechanism to gauge buyer default risk (see, for example, Mian and Smith, 1992, and Frank and Maksimovic, 2005). In particular, sellers can reduce payment risks through two-part payment terms, such as early payment discounts (e.g., Ng et al., 1999). Alternatively, suppliers with high borrowing costs might offer an early payment discount to reduce the need to finance their own extension of trade credit to buyers. In sum then, we see four important, and not mutually exclusive, reasons for trade credit: 1) As a method of financing; 2) As a means of price discrimination or market power; 3) As a bond assuring buyers of product quality; 4) As a screening mechanism to gauge buyer default risk. These four reasons offer several testable hypotheses, not all of which are mutually exclusive. Hypothesis 1: Suppliers in industries with substantial turnover and perishable goods extend shorter net days. This is consistent with both the financing and bonding explanations. Hypothesis 2: Smaller suppliers extend longer net days. This is consistent with both a market power explanation (small suppliers are squeezed more by buyers) and a bonding explanation (small suppliers have to offer better bonds). Hypothesis 3: Larger suppliers extend longer net days. This is consistent with a financing explanation (large suppliers have lower financing costs). 10 Hypothesis 4: Investment grade suppliers extend longer net days. This is consistent with a financing explanation (the cost of finance is less for investment grade suppliers, allowing them to offer longer terms), a buyer market power explanation (suppliers for whom providing credit costs less may be squeezed for more), and a bonding explanation (in order to signal commitment to quality, an investment grade supplier who can raise finance at lower cost will offer longer terms). Hypothesis 5: Large buyers receive longer net days. This is consistent with both a market power explanation (suppliers are squeezed more by large buyers) and a screening explanation (large buyers have lower default risk and need not be screened as much). Hypothesis 6: Investment grade buyers receive longer net days. This is consistent with a screening explanation (investment-grade buyers have lower default risk and need not be screened as much). Hypothesis 7: Discounts are more common for small and non-investment grade buyers. This is consistent with the screening explanation (early payment discounts offer suppliers a screening mechanism to gauge buyer default risk). Hypothesis 8: Discounts are more common from small suppliers. This is consistent with the market power hypothesis (large buyers demand discounts from small suppliers) and the screening explanation (it is more difficult for small suppliers to absorb and diversify default risk). Table 5 summarizes these testable hypotheses for different trade credit terms: net days versus discount offered. Panel A summarizes the role of supplier and buyer characteristics in explaining differences in contract maturity as measured by the net days, and shows whether under each of the four different explanations for trade credit the effect on net days is positive, negative or zero. Panel B summarizes the role of supplier and buyer characteristics in explaining whether or not early payment discounts are offered, and shows whether under each of the four 11 different explanations for trade credit the effect on the likelihood of discounts is positive, negative, or zero. With these eight possible explanations and testable hypotheses in mind, let us examine the data more carefully. IV. Regression Analysis In this section, we use a multivariate framework to study the determinants of contract terms. Summary statistics and definitions of all variables are shown in Table 6. Our first dependent variable is the contract maturity, measured as the log number of net days. The strict financing explanation would suggest that large, investment grade suppliers (who have easier access to credit) should offer longer terms to small, non-investment grade buyers. The price discrimination cum bargaining power explanation would suggest that larger buyers should obtain more credit from smaller suppliers. The view that trade credit is posted as a bond assuring product quality would suggest that smaller suppliers (who typically have less of a history and reputation) should offer longer terms, while buyer size should not matter. We include supplier and buyer characteristics as explanatory variables. We include an indicator if the buyer is big (above $ 10 billion in sales), as well as an indicator if the buyer has an investment grade rating. Similarly, we include an indicator if the supplier is large (above $ 2 billion in sales), an indicator if the supplier is medium sized (between $ 100 million and $ 2 billion in sales), as well as an indicator if the supplier is investment grade. We also include indicators for the buyer's industry. Correlation matrices of all variables are shown for the full sample and subsample of contracts that offer a pre-payment discount in Table 7. Although there are significant relationships between our explanatory variables, the correlation levels are generally sufficiently low to eliminate concerns of cross-correlation among variables. Our first results are shown in Table 8. The first two columns cluster standard errors by buyer, while the next two columns include buyer fixed effects, and the last two columns include supplier fixed effects. The second columns in each of these pairs excludes credit contracts with 12 discounts to abstract from the possibility that net days on two-part contracts vary systematically from those of simple contracts without discounts. Our industry classifications are very broad. Nevertheless, we find buyers in industries with substantial turnover (groceries, soft goods), and where goods are more likely to be perishable, tend to have shorter net days.8 This is consistent with both the financing and bonding explanations. Perhaps most interestingly, we find that longer net days are offered to significantly larger, investment grade buyers (Table 8 Columns 1-2 and Columns 3-4). The magnitude of these effects is sizeable. For example, from the estimates in Column 2 a buyer who is large gets 9.8 longer days than the mean of 59 days. Similarly, a buyer who is investment grade gets 7.5 longer days than the mean net days. These results empirically support the hypothesis that trade credit terms are used as a means of price discrimination. We also find that net days are shorter for buyers located in North America (the majority of which are located in the US) relative to buyers located in Europe. One potential explanation for this result is that sales in Europe are often cross-border in which case buyers may demand longer days to protect against damaged goods and avoid having to challenge suppliers in foreign courts. When we include supplier fixed effects (thus focusing on the subsample of suppliers with multiple contracts within or across buyers), we continue to find that larger and investment grade buyers get longer net days (Table 8, Columns 5 and 6). These regressions exclude observations from suppliers without multiple contracts. When we include buyer fixed effects (Table 8, Columns 3 and 4), we find that longer net days are significantly more likely to be extended by smaller suppliers. This is consistent with both a market power explanation (small suppliers are squeezed more by buyers) and a bonding explanation (small suppliers, ceteris paribus, have to offer better bonds). Finally, we also find that investment grade suppliers extend longer net days. This is consistent with any of the first three explanations above ­ clearly, it costs investment grade suppliers less to provide a given amount of financing, so in any financing explanation, 8 We do not attach much importance to the industry effect found for the utility sector because it is based on observations from only one firm. 13 investment grade suppliers will offer longer terms. For any level of buyer market power, one could argue that a supplier for whom providing credit costs less will be squeezed for more. Similarly, in order to signal commitment to quality, an investment grade supplier who can raise finance at lower cost will offer longer terms. Next, we examine the sample of contracts that include an early payment discount. The view that discounts are used as a screening mechanism to gauge buyer default risk would suggest that smaller and non-investment grade buyers, where default risk tends to be higher, would more likely receive discounts. To the extent that it is easier to absorb and diversify default risk for large firms, this view would also suggest that small suppliers are more likely to extend discounts. On the other hand, if there are fixed costs in screening, one would expect that large suppliers are more likely to offer discounts. Table 9 shows logit regressions of determinants of early payment discounts for the subsample of contracts that offer early payment discounts. The dependent variable takes value 1 if the contract includes a discount (two-part contract), and 0 otherwise. As before, the first two columns present results for regressions with buyer clustered standard errors, the next two columns present results for regressions with buyer fixed effects, and the final two columns present results for regressions with supplier fixed effects. In the second of each of the regression pairs, we drop observations from buyers that never receive discounts, to abstract from the possibility that such firms are systematically different from firms that receive discounts. We find that discounts are less common for large and investment grade buyers, consistent with the screening hypothesis according to which early payment discounts offer suppliers a screening mechanism to gauge buyer default risk. The buyer fixed effects regressions in columns 3 and 4 indicate that early payment discounts are more common from small suppliers. This is consistent with the market power hypothesis according to which large buyers demand discounts from small suppliers, despite the fact that two-part contracts are generally considered to be more expensive to administer. This result is also in line with Burkart, Ellingsen, and Giannetti (2008) who find using U.S. survey data that firms that have more buyer market power receive larger early payment discounts. This finding is also consistent with the screening view that stipulates that smaller suppliers are more 14 likely to offer discounts as a screening mechanism for nonpayment default because it is more difficult for these firms to absorb and diversify default risk. The supplier fixed effects regressions in columns 5 and 6 confirm that suppliers are more likely to offer early payment discounts to smaller and non-investment grade buyers, consistent with the screening hypothesis. Furthermore, these regressions display strong industry effects. It should be noted that these regressions are based on a relatively small sample of suppliers because only 85% of the suppliers with multiple contracts display variation in whether or not their contracts include early payment discounts, indicating a strong supplier fixed effect in whether or not firms extend early payment discounts. Finally, we analyze the determinants of discount terms for the subsample of contracts that offer early payment discounts and for which we have complete information on discount terms (including discount period, discount rate, and net days). The results are not materially affected when we include firms with incomplete information on discount terms. Discount terms appear to be strongly dependent on industry norms. For instance, buyers of soft goods and groceries tend to receive the longest discount days. The same industries also receive the highest effective rates on two-part contracts. It is also worth emphasizing that a surprisingly large fraction of contracts (over 30%) with early payment discounts have a spread between net days and discount days of exactly one day, suggesting that discounts are often used to encourage prompt payments. In Table 10, Columns 1 to 3, we regress our buyer and supplier characteristics on the ratio of discount days/net days, for the subsample of contracts that offer early payment discounts. Columns 4 and 5 of Table 10 show results from a multinomial logistic regression model, which includes the full sample of contracts: the dependent variable is equal to zero if the contract does not include an early payment discount; equal to one if net days less discount days is greater than one; and equal to two if net days less discount days is equal to zero or one. In all specifications, we find a significant relationship with investment grade buyers, contrary to our finding in Table 9 that investment grade buyers are less likely, overall, to receive early payment discounts. In the multinomial logit regression, the coefficients on investment 15 grade buyers are significant in both columns (denoting net days less discount days is greater than one or net days less discount days is equal to zero or one, respectively), but alter signs between specifications. Investment grade buyers are less likely to receive discounts with net days in excess of discount days, and more likely to receive discounts with discount days close to net days, suggesting that discounts are often used to encourage prompt payments from investment grade buyers. We also find that buyers in North America are significantly more likely to be offered discount days equal to net days. In addition, supplier characteristics are no longer significant. This may suggest that while suppliers in Europe use early payment discounts as a risk management tool to encourage early payment from riskier buyers, suppliers in North America use discounts as another competitive gesture, or `sweetener', in alternative or in addition to up- front discounts. In unreported regressions, we generally find similar patterns across discount terms (including discount period, discount rate, and the effective discount rate9) in the sense that the coefficients on the various firm determinants have the same sign in most specifications, suggesting that the different discount terms serve similar purposes and that firms do not systematically trade off various terms against each other. This is consistent with the findings by Ng et al. (1999). For example, while grocers tend to receive longer discount days relative to net days, they also pay higher effective rates. Similarly, investment grade buyers tend to both receive longer discount days and pay higher effective rates, conditional upon receiving a discount. However, these regressions are hard to interpret because some of the same firm characteristics that determine whether or not firms receive early prepayment discounts also appear to affect discount terms. Overall, we find that trade credit contract terms are jointly determined by supplier and buyer characteristics, based on explanations of market power, information asymmetries, and alternative financing costs. 9 The effective discount rate is computed as (1/ (1 ­ discount rate))360 / (net days ­ discount days) ­ 1).). 16 V. Concluding Remarks This paper provides new evidence on the unique role of trade credit as a competitive gesture and risk management tool used by suppliers and as a potentially inexpensive source of working capital financing for large buyers. We use a novel dataset on almost 30,000 supplier contracts for 56 large buyers in Europe and North America to document contract terms across buyers and suppliers of varying size and investment quality. The bilateral, multi-contract nature of our dataset is truly unique and is a major improvement on (generally survey based) datasets that have previously been used to study the determinants of credit terms used in trade credit. This multi-contract structure of our dataset allows us to abstract from unobserved buyer and supplier firm characteristics, something previous empirical work has not been able to do. We find that the largest and most creditworthy buyers receive contracts with the longest maturities, as measured by net days, from smaller, investment grade suppliers, consistent with existing trade credit theories. In particular, these results are consistent with a market power explanation (smaller suppliers are squeezed more by large buyers), a screening explanation (large and investment grade buyers have lower default risk and need not be screened as much), a bonding explanation (small suppliers have to offer better bonds), and a financing explanation (investment grade suppliers have lower financing costs, allowing them to offer longer terms). In comparison, early payment discounts seem to be used as a risk management tool to limit the potential nonpayment risk of trade credit. In particular, early payment discounts are generally offered by smaller, non-investment grade suppliers to smaller and non-investment grade buyers. This is consistent with the screening explanation (early payment discounts offer suppliers a screening mechanism to gauge buyer default risk). Our result that discounts are more common from small suppliers is also consistent with a market power explanation, whereby large buyers demand discounts from small suppliers. Our results suggest that contract terms are jointly determined by supplier and buyer characteristics, based on market power, information asymmetries, and alternative financing costs. 17 References Bates, Thomas, Kathleen Kahle, and Rene Stulz, 2008. Why do U.S. Firms Hold So Much More Cash Than They Used To?, Journal of Finance, forthcoming. Biais, Bruno and Christian Gollier, 1997. Trade Credit and Credit Rationing. Review of Financial Studies 10, 903­937. Boissay, Frederic and Reint Gropp, 2007. Trade Credit Defaults and Liquidity Provision by Firms, European Central Bank Working Paper no. 753. Brennan, Michael, Vojislav Maksimovic and Josef Zechner, 1988. Vendor Financing. Journal of Finance 43, 1127­1141. Burkart, Mike and Tore Ellingsen, 2004. In-Kind Finance: A Theory of Trade Credit. American Economic Review 94, 569­590. Burkart, Mike, Tore Ellingsen, Mariassunta Giannetti, 2008. What You Sell is What You Lend? Explaining Trade Credit Contracts. Review of Financial Studies, forthcoming. Choi, Woon Gyu and Yungsan Kim, 2005. Trade Credit and the Effect of Macro-Financial Shocks: Evidence from U.S. Panel Data, Journal of Financial and Quantitative Analysis 40, 897- 925. Cunat, Vicente, 2007. Trade Credit: Suppliers and Debt Collectors as Insurance Providers, Review of Financial Studies 20, 491­527. Demirguc-Kunt, Asli, and Vojislav Maksimovic, 1998. Law, Finance, and Firm Growth. Journal of Finance 6, 2107­2137. Demirguc-Kunt, Asli, and Vojislav Maksimovic, 1999. Institutions, Financial Markets, and Firms Debt Maturity. Journal of Financial Economics 54, 295­336. Demirguc-Kunt, Asli, and Vojislav Maksimovic, 2002. Firms as Financial Intermediaries: Evidence from Trade Credit Data, World Bank Working Paper. Fabbri, Daniela and Leora Klapper, 2008. Market Power and the Matching of Trade Credit Terms. Working Paper. University of Amsterdam. Ferris, J. Stephen, 1981. A Transaction Theory of Trade Credit Use. The Quarterly Journal of Economics 96, 247­270. Fisman, Raymond and Inessa Love, 2003. Trade Credit, Financial Intermediary Development, and Industry Growth, Journal of Finance 58, 353­374. 18 Fisman, Raymond, and Mayank Raturi, 2004. Does Competition Encourage Credit Provision? Evidence from African Trade Credit Relationships. Review of Economics and Statistics 86, 345- 352. Frank, Murray and Vojislav Maksimovic, 2005. Trade Credit, Collateral, and Adverse Selection. University of Maryland, Working Paper. Johnson, Simon, John McMillan and Christopher Woodruff, 2002. Courts and Relational Contracts. Journal of Law, Economics and Organization 18, 221­277. Lee, Yul W. and John D. Stowe, 1993. Product Risk, Asymmetric Information, and Trade Credit. Journal of Financial and Quantitative Analysis 28, 285­300. Long, Michael S., Ileen B. Malitz, and S. Abraham Ravid, 1993. Trade Credit, Quality Guarantees, and Product Marketability. Financial Management 22, 117­127. Love, Inessa, Lorenzo A. Preve, and Virginia Sartia-Allende, 2007. Trade Credit and Bank Credit: Evidence from Recent Financial Crises. Journal of Financial Economics 83, 453­ 469. Marotta, Giuseppe, 2005. Is Trade Credit More Expensive than Bank Credit Loans? Evidence from Italian Firm-Level Data. Applied Economics 37, 403­416. McMillan, John and Christopher Woodruff, 1999. Interfirm Relationships and Informal Credit in Vietnam, Quarterly Journal of Economics 114, 1285­1320. Mian, Shehzad L. and Clifford W. Smith, 1992. Accounts Receivable Management Policy: Theory and Evidence. Journal of Finance 47, 169­200. Ng, Chee K., Janet K. Smith and Richard L. Smith, 1999, Evidence on the Determinants of Credit Terms Used in Interfirm Trade. Journal of Finance 54, 1109­1129. Petersen, Mitchell and Raghuram G. Rajan, 1995. The Effect of Credit Market Competition on Lending Relationships. Quarterly Journal of Economics 110, 407­443. Petersen, Mitchell A. and Raghuram G. Rajan, 1997. Trade Credit: Theory and Evidence. Review of Financial Studies 10, 661­691. Schiff, Michael and Zvi Lieber, 1974. A Model for the Integration of Credit and Inventory Management, Journal of Finance 29, 133-140. Smith, Janet K., 1987. Trade Credit and Informational Asymmetry. Journal of Finance 42, 863­ 872. Van Horen, Neeltje, 2005. Do Firms Use Trade Credit as a Competitiveness Tool? Evidence from Developing Countries, World Bank Working Paper. 19 Wilner, Benjamin S. 2000. The Exploitation of Relationships in Financial Distress: The Case of Trade Credit. Journal of Finance 55, 153­178. 20 Table 1: Buyer and Seller Characteristics This table reports summary statistics of buyer and supplier characteristics. Sample consists of 29,019 trade credit contracts between 56 large buyers and 24,140 suppliers. Panel A: Buyer Characteristics Number of % of Number % of Total Spend % of total Buyers buyers Contracts Contracts ($mln) spend Size >$10B 33 59 24,298 84 612 89 Size $0.1- 10B 23 41 4,721 16 79 11 Industry auto 9 16 1,615 6 75.8 11 Industry diversified retail 7 13 9,749 34 193 28 Industry diversified mfg 16 29 3,824 13 74.3 11 Industry grocery 4 7 1,630 6 88.5 13 Industry hard goods retail 9 16 3,146 11 164 24 Industry soft goods retail 6 11 2,362 8 42.1 6 Industry technology 3 5 5,306 18 24.2 4 Industry food & beverages 2 4 682 2 26.7 4 Industry utility 1 2 705 2 2.47 0 Location: Europe 13 23 12,029 41 241 35 Location: North America 43 77 16,990 59 450 65 Investment Grade: No 14 25 4,514 16 42.9 7 Investment Grade: Yes 42 75 24,505 84 570 93 Panel B: Supplier Characteristics: Number of % of Number % of Total Spend % of total Suppliers Suppliers Contracts Contracts ($mln) spend Size >$2B 2,727 11 5,772 20 531 77 Size $0.1-2B 7,821 32 9,549 33 142 21 Size <$0.1B 13,590 56 13,698 47 17.9 3 Investment Grade: No 16,391 68 18,655 65 319 46 Investment Grade: Yes 7,713 32 10,043 35 372 54 21 Table 2: Contract Characteristics This table reports summary statistics of trade credit contract characteristics. Sample consists of 29,019 trade credit contracts between 56 large buyers and 24,140 suppliers. Early payment discounts are offered on 3,717 of these contracts. N Mean Median Min Max Std Dev Panel A: All Contracts: Spend (millions) 29,019 23.8 3.47 .0004 6,520 111.0 Spend $1M 29,019 0.28 0 0 1 Spend >$1M - $4M 29,019 0.25 0 0 1 Spend >$4M - $15M 29,019 0.22 0 0 1 Spend >$15M 29,019 0.25 0 0 1 Net Days 29,019 59.2 60 1 120 26.1 Net Days 0 ­ 30 29,019 0.25 0 0 1 Net Days 31 ­ 60 29,019 0.37 0 0 1 Net Days 61 ­ 90 29,019 0.24 0 0 1 Net Days >90 29,019 0.11 0 0 1 Discount offered 3,717 0.13 0 0 1 Panel B: Subsample of Contracts that offer an early payment discount: Discount Days 3,462 30.43 30 1 180 20.09 Discount Days 0 ­ 30 3,462 0.64 1 0 1 Discount Days 31 ­ 60 3,462 0.27 0 0 1 Discount Days >60 3,462 0.09 0 0 1 Discount rate (%) 3,707 2 2 .02 11.5 0.09 Discount 1% 3,707 36 0 0 1 Discount >1% - 2% 3,707 56 1 0 1 Discount > 2% 3,707 8 0 0 1 Ratio of Discount to Net Days 2,634 0.63 0.6 0.02 1 0.28 Effective Interest Rate 2,584 0.53 0.27 0.02 1 0.38 22 Table 3: Distribution of Buyer and Seller Characteristics by Contract Characteristics This table reports the distribution (in percentages) of trade credit contract terms by buyer and supplier characteristics. Sample consists of 29,019 trade credit contracts between 56 large buyers and 24,140 suppliers. NA denotes North America. Spend (%) Net Days (%) Location < $1M $1-4M >$4-15M > $15M 0-30 31-60 61-90 91+ Europe NA Buyer Characteristics: Size >$10B 32 23 20 25 20 38 29 13 46 54 Size $0.1- 10B 5 35 33 26 52 40 8 0 18 82 Industry auto 0 15 28 57 20 53 20 7 59 41 Industry diversified retail 23 37 22 17 10 13 50 26 84 16 Industry diversified mfg 6 29 36 29 47 34 17 1 0 100 Industry grocery 11 25 18 47 42 54 3 0 84 16 Industry hard goods retail 0 15 29 57 21 52 24 4 3 97 Industry soft goods retail 26 23 22 29 85 14 1 1 0 100 Industry technology 74 9 10 7 8 88 3 0 0 100 Industry food & beverages 28 34 21 17 27 30 12 31 100 0 Industry utility 73 18 6 4 54 9 36 0 100 0 Location: Europe 26 33 20 20 11 19 46 23 100 0 Location: North America 29 19 23 29 37 53 10 1 0 100 Investment Grade: No 18 19 31 33 67 30 2 0 48 52 Investment Grade: Yes 30 26 20 24 19 40 29 12 6 94 Supplier Characteristics: Size: >$2B 5 11 18 65 33 39 21 8 30 70 Size: $0.1-2B 5 89 48 38 34 32 24 11 39 61 Size: <$0.1B 53 41 6 0 18 43 27 12 48 52 Investment Grade: No 28 26 23 23 27 40 24 10 43 57 Investment Grade: Yes 27 22 21 30 23 37 27 12 41 59 23 Table 4: Distribution of Buyer and Seller Characteristics by Discount Characteristics This table reports the distribution of trade credit contract terms by buyer and supplier characteristics for the subsample of 3,717 contracts that offer an early payment discount. Full Subsample of Contracts that Offer an Early Payment Discount Sample Discount Rate (%) Discount Days (%) Discount to Discount Net days 0-1% 1-2% > 2% 0-30 31-60 61+ (%) Ratio (%) Buyer Characteristics: Size >$10B 10 35 58 7 64 33 3 64 Size $0.1- 10B 26 37 52 11 78 21 2 60 Industry auto 19 21 50 29 100 0 0 35 Industry diversified retail 5 34 66 0 94 5 1 43 Industry diversified mfg 13 67 30 3 95 4 1 44 Industry grocery 25 31 48 22 84 15 0 87 Industry hard goods retail 58 35 60 5 54 42 4 68 Industry soft goods retail 8 5 86 9 25 75 0 95 Industry technology 0 . . . . . . . Industry food & beverages 0 . . . . . . . Industry utility 0 . . . . . . . Location: Europe 4 19 44 37 70 28 2 80 Location: North America 19 38 58 4 67 30 3 61 Investment Grade: No 25 37 60 3 76 21 2 54 Investment Grade: Yes 11 35 54 11 64 33 3 66 Supplier Characteristics: Size >$2B 27 34 58 7 66 31 3 65 Size $0.1-2B 17 38 54 8 67 30 2 63 Size <$0.1B 4 32 58 11 80 19 1 57 Investment Grade: No 13 37 55 9 69 29 2 63 Investment Grade: Yes 12 33 59 7 67 30 3 64 24 Table 5: Testable Hypotheses Based on Trade Credit Theories This table summarizes the testable hypotheses for trade credit terms (net days and discount offered) that follow from the four, not mutually exclusive explanations for trade credit offered in the text: as a method of financing; as a means of price discrimination or market power; as a bond assuring buyers of product quality; or as a screening mechanism to gauge buyer default risk. Panel A summarizes the role of supplier and buyer characteristics in explaining differences in contract maturity as measured by net days. In this panel, - denotes shorter net days; + denotes longer net days; and 0 denotes no effect on net days. Panel B summarizes the role of supplier and buyer characteristics in explaining whether or not early payment discounts are offered. In this panel, - denotes discounts less common; + denotes discounts more common; and 0 denotes no effect on likelihood of discounts. Panel A: Net days Financing Market Power Bonding Screening Supplier Characteristics: High turnover/perishable goods - 0 - 0 Small in size - + + 0 Investment grade + + + 0 Buyer Characteristics: Large in size 0 + 0 + Investment grade 0 0 0 + Panel B: Early payment discount Financing Market Power Bonding Screening Supplier Characteristics: Small in size 0 + 0 + Buyer Characteristics: Small in size 0 0 0 + Non-Investment grade 0 0 0 + 25 Table 6: Summary Statistics of Regression Variables This table reports summary statistics of the main regression variables. Sample consists of 29,019 trade credit contracts between 56 large buyers and 24,140 suppliers. Early payment discounts are offered on 3,717 of these contracts. Variable Obs Mean Std. Dev. Min Max Complete sample: Log net days 28,187 3.9468 0.6138 0 5.4806 Discount dummy 29,019 0.1281 0.3342 0 1 Subsample of contracts with early payment discount: Discount days 3,462 30.4330 20.0922 1 180 Discount rate 3,707 0.0173 0.0084 0.0002 0.1150 Discount days/Net days 2,634 0.6299 0.2815 0.0167 1 Effective rate 2,584 0.5332 0.3813 0.0169 1 Buyer characteristics: Buyer large size 29,019 0.8373 0.3691 0 1 Buyer small size 29,019 0.1627 0.3691 0 1 Buyer investment grade 29,019 0.8444 0.3624 0 1 Buyer North America 29,019 0.5855 0.4926 0 1 Industry auto 29,019 0.0557 0.2293 0 1 Industry diversified retail 29,019 0.3360 0.4723 0 1 Industry diversified mfg 29,019 0.1318 0.3383 0 1 Industry grocery 29,019 0.0562 0.2303 0 1 Industry hard goods retail 29,019 0.1084 0.3109 0 1 Industry soft goods retail 29,019 0.0814 0.2734 0 1 Industry technology 29,019 0.1828 0.3865 0 1 Industry food and beverages 29,019 0.0235 0.1515 0 1 Industry utility 29,019 0.0243 0.1540 0 1 Supplier characteristics: Supplier large size 29,019 0.1989 0.3992 0 1 Supplier medium size 29,019 0.3291 0.4699 0 1 Supplier small size 29,019 0.4720 0.4992 0 1 Supplier investment grade 29,019 0.3461 0.4757 0 1 26 Table 7: Correlation Matrix of Regression Variables This table reports correlations between the main regression variables. Panel A presents correlations between the dependent variables; Panel B presents correlations between the explanatory and dependent variables; and Panel C presents correlations between explanatory variables. Sample consists of 29,019 trade credit contracts between 56 large buyers and 24,140 suppliers. Early payment discounts are offered on 3,717 of these contracts. Panel A: Dependent Variables Log net days Discount days/net days Discount days/net days -0.19* Effective rate -0.41* 0.32* Panel B: Explanatory and Dependent Variables Full Sample Subsample w/Discount Discount Effective Log net Discount days/net rate days dummy days Buyer large size 0.26* -0.18* 0.09* -0.10 Buyer small size -0.26* 0.18* -0.09* 0.06 Buyer investment grade 0.30* -0.15* 0.20* 0.13* Buyer North America -0.37* 0.22* -0.18* -0.32* Industry auto 0.00 0.04* -0.17* -0.00 Industry diversified retail 0.35* -0.17* -0.33* -0.10 Industry diversified mfg -0.13* 0 -0.32* -0.00 Industry grocery -0.17* 0.09* 0.35* 0.33* Industry hard goods retail -0.00 0.47* 0.15* -0.08* Industry soft goods retail -0.25* -0.04* 0.33* 0.11* Industry technology -0.02* -0.18* . . Industry food and beverage 0.05* -0.06* . . Industry utility -0.07* -0.06* . . Supplier large size -0.11* 0.21* 0.06* 0.10* Supplier medium size -0.10* 0.08* 0.00 -0.12* Supplier small size 0.18* -0.25* -0.08* 0.04 Supplier investment grade 0.03* -0.02* 0.02 0.04 Note: Asterisks indicate significance at 1% 27 Panel C: Explanatory Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) Buyer large size (1) 1.00 Buyer small size (2) -1.00* 1.00 Buyer investment grade (3) -0.00 0.00 1.00 North America (4) -0.21* 0.21* -0.31* 1.00 Ind: auto (5) -0.15* 0.15* -0.10* -0.08* 1.00 Ind: diversified retail (6) 0.17* -0.17* 0.17* -0.62* -0.17* 1.00 Ind: diversified mfg(7) -0.38* 0.38* -0.06* 0.33* -0.09* -0.28* 1.00 Ind: grocery (8) -0.70* 0.70* -0.01 0.17* 0.26* -0.08* 0.37* 1.00 Ind: hard goods retail (9) -0.03* 0.03* -0.04* 0.27* -0.08* -0.25* -0.14* -0.11* 1.00 Ind: soft goods retail (10) -0.07* 0.07* -0.32* 0.25* -0.07* -0.21* -0.12* -0.09* -0.10* 1.00 Ind: technology (11) 0.21* -0.21* 0.15* 0.40* -0.11* -0.34* -0.18* -0.15* -0.16* -0.14* 1.00 Ind: food&beverages (12) -0.03* 0.03* -0.04* -0.18* -0.04* -0.11* -0.06* -0.05* -0.05* -0.05* -0.07* 1.00 Ind: utility (13) 0.07* -0.07* 0.07* -0.19* -0.04* -0.11* -0.06* -0.05* -0.06* -0.05* -0.07* -0.02* 1.00 Supplier large size (14) -0.00 0.00 -0.12* 0.12* 0.07* -0.10* 0.03* -0.01 0.23* -0.03* -0.15* 0.00 -0.06* 1.00 Supplier medium size (15) -0.12* 0.12* -0.09* 0.04* 0.11* -0.05* 0.13* 0.02* 0.10* 0.04* -0.20* -0.01 -0.07* -0.35* 1.00 Supplier small size (16) 0.12* -0.12* 0.18* -0.13* -0.16* 0.12* -0.14* -0.01 -0.28* -0.01 0.31* 0.01 0.11* -0.47* -0.66* 1.00 Supplier investment grade (17) 0.08* -0.08* 0.00 -0.02* 0.02* 0.02* -0.03* -0.07* 0.02* -0.05* 0.03* -0.02* 0.00 0.22* -0.11* -0.08* Note: * denotes statistical significance at the 1% level. 28 Table 8: Log net days Dependent variable is the logarithm of net days on the contract. Standard errors in regressions (1) and (2) are corrected for clustering at the buyer level. Regressions (3) and (4) include buyer fixed effects. Regressions (5) and (6) include supplier fixed effects and are estimated based on the subsample of suppliers that have multiple contracts. Regressions (2), (4) and (6) include only trade credit contracts without discounts. Standard errors are reported between brackets. ***, **, * indicate significance at the 1%, 5%, and 10% level, respectively. Dependent variable: Log net days Buyer clustered Buyer FE Supplier FE Without discount Without discount Without discount (1) (2) (3) (4) (5) (6) Buyer large size 0.389*** 0.391*** 0.225*** 0.170*** (0.102) (0.110) (0.025) (0.031) Buyer investment grade 0.242** 0.295*** 0.217*** 0.260*** (0.102) (0.110) (0.024) (0.028) Buyer North America -0.485*** -0.554*** -0.446*** -0.502*** (0.167) (0.184) (0.043) (0.054) Industry diversified retail -0.072 -0.136 0.042 -0.053 (0.127) (0.155) (0.060) (0.071) Industry diversified mfg 0.094 0.123 0.183*** 0.173*** (0.165) (0.185) (0.052) (0.060) Industry grocery -0.764*** -0.760*** -0.529*** -0.521*** (0.136) (0.163) (0.069) (0.084) Industry hard goods retail 0.140 0.126 0.084 0.031 (0.146) (0.162) (0.060) (0.072) Industry soft goods retail -0.263* -0.230 -0.118 -0.091 (0.143) (0.161) (0.086) (0.097) Industry technology -0.051 -0.047 -0.231*** -0.277*** (0.304) (0.321) (0.062) (0.069) Industry food and beverage -0.148 -0.199 0.090 0.033 (0.142) (0.157) (0.131) (0.141) Industry utility -0.806*** -0.872*** -0.551*** -0.652*** (0.122) (0.156) (0.119) (0.128) Supplier large size -0.148 -0.143 -0.059*** -0.068*** (0.126) (0.138) (0.008) (0.008) Supplier medium size -0.140 -0.155 -0.041*** -0.047*** (0.124) (0.134) (0.006) (0.006) Supplier investment grade 0.008 0.006 0.017*** 0.017*** (0.016) (0.017) (0.005) (0.005) Number of buyers 56 56 56 56 56 56 Number of suppliers 24,006 22,028 24,006 22,028 2,267 2,051 Number of observations 28,187 25,298 28,187 25,298 6,448 5,321 R-squared 0.337 0.335 0.036 0.030 0.276 0.286 29 Table 9: Discounts Dependent variable is a dummy variable that take a value of one if the trade credit contract includes a discount (two-part contract), and zero otherwise. Regression estimates are based on a logit model. Standard errors in regressions (1) through (2) are corrected for clustering at the buyer level. Regressions (3) and (4) include buyer fixed effects. Regressions (5) through (6) include supplier fixed effects and are estimated based on the subsample of suppliers that have multiple contracts. Regressions (2), (4) and (6) exclude buyers with no discounts ever. Several industries do not have firms with discounts and are dropped from estimation. Standard errors are reported between brackets. ***, **, * indicate significance at the 1%, 5%, and 10% level, respectively. Dependent variable: Discount Buyer clustered Buyer FE Supplier FE w/o no discount w/o no discount w/o no discount (1) (2) (3) (4) (5) (6) Buyer large size -1.653** -0.854* -1.650*** -1.269*** (0.686) (0.502) (0.174) (0.190) Buyer investment grade -0.814 -0.888* -0.668*** -0.651*** (0.517) (0.535) (0.139) (0.157) Buyer North America 1.666 0.716 3.318*** 3.528*** (1.274) (1.141) (0.378) (0.509) Industry diversified retail 0.922 -0.752 2.452*** 0.624 (1.377) (1.034) (0.415) (0.512) Industry diversified mfg -0.010 -1.619* -0.157 -1.471*** (1.080) (0.955) (0.334) (0.460) Industry grocery 3.191* -0.073 5.520*** 3.182*** (1.655) (1.118) (0.546) (0.678) Industry hard goods retail 2.916*** 0.556 3.637*** 1.417*** (0.910) (1.049) (0.371) (0.466) Industry soft goods retail -0.086 -1.184 0.575 -0.592 (1.136) (1.007) (0.750) (0.850) Supplier large size 1.147*** 0.553 -0.225** -0.225** (0.340) (0.415) (0.095) (0.095) Supplier medium size 0.697** 0.153 -0.322*** -0.322*** (0.286) (0.329) (0.083) (0.083) Supplier investment grade -0.279*** -0.097 -0.065 -0.065 (0.094) (0.107) (0.064) (0.064) Number of buyers 56 34 34 34 56 34 Number of suppliers 24,140 7,927 7,927 7,927 399 305 Number of observations 29,019 10,604 10,604 10,604 2,067 1,433 30 Table 10: Discount Terms Dependent variable is the ratio of discount days to net days in regressions (1)-(3). Columns (4) and (5) report results of a multinomial logit regression where the dependent variable takes a value of 2 if net days minus discount days is less than or equal to 1, a value of 1 if net days minus discount days is more than 1, and a value of 0 if the contract offers no discount (which we set as the base outcome). Standard errors in regression (1) and the multinomial regression reported in columns (4) and (5) are corrected for clustering at the buyer level. Regression (2) includes buyer fixed effects and regression (3) includes supplier fixed effects and is estimated based on the subsample of suppliers that have multiple contracts. Several industries do not have firms with discounts and are dropped from estimation in regressions (1) to (3). We also exclude contracts with missing discount or net days information from the regressions in (1) to (3). Standard errors are reported between brackets. ***, **, * indicate significance at the 1%, 5%, and 10% level, respectively. Dependent variable: Discount days/Net days Multinomial logit Net days minus Net days minus discount days>1 discount days<=1 Buyer clustered Buyer FE Supplier FE Buyer clustered Buyer clustered (1) (2) (3) (4) (5) Buyer large size 0.012 0.002 -1.383* -1.787** (0.112) (0.030) (0.741) (0.814) Buyer investment grade 0.082* 0.148*** -1.139** 2.578*** (0.049) (0.033) (0.555) (0.993) Buyer North America 0.220*** 0.380*** 1.369 5.768*** (0.048) (0.078) (1.371) (1.508) Industry diversified retail 0.074 -0.157 0.819 4.318*** (0.058) (0.101) (1.398) (1.609) Industry diversified mfg 0.049 -0.081 0.128 2.770* (0.072) (0.091) (1.081) (1.532) Industry grocery 0.611*** 0.302*** 1.683 12.095*** (0.057) (0.105) (1.778) (1.999) Industry hard goods retail 0.261*** -0.150 2.786*** 6.765*** (0.093) (0.099) (0.943) (1.351) Industry soft goods retail 0.539*** 0.331* -3.911*** 5.636*** (0.070) (0.174) (1.130) (1.503) Supplier large size -0.016 0.010 0.993*** 0.980 (0.032) (0.013) (0.336) (0.603) Supplier medium size -0.026 0.010 0.563* 0.509 (0.026) (0.012) (0.309) (0.466) Supplier investment grade 0.006 0.001 -0.262** -0.206 (0.017) (0.009) (0.105) (0.132) Number of buyers 34 34 28 56 Number of suppliers 2,080 2,080 531 24,140 Number of observations 2,584 2,584 1,035 29,019 R-squared 0.440 0.008 0.254 0.361 31 Appendix: Variable Definitions Variable Name Variable Definition Net days Natural logarithm of net days of the contract in number of days. Discount days Natural logarithm of discount days of the contract in number of days. Discount rate Discount percentage Discount days / Net days Ratio of number of discount days to number of net days Effective rate Effective rate computed as 1 (1 discount rate) 360 /( net days discount days ) 1 Buyer large size Dummy variable that takes a value of one when the total sales of the buyer exceeds US$ 10 billion, and zero otherwise. Buyer small size Dummy variable that takes a value of one when the total sales of the buyer is smaller than US$ 10 billion, and zero otherwise. Buyer investment grade Dummy variable that takes a value of one when the buyer is investment grade, and zero otherwise. Buyer North America Dummy variable that takes a value of one when the location of the buyer is North America (Canada or US) and zero when the location of the buyer is elsewhere (UK or Continental Europe). Industry auto Dummy variable that takes a value of one if the buyer is active in the auto industry, and zero otherwise. Industry diversified retail Dummy variable that takes a value of one if the buyer is active in the diversified retail industry, and zero otherwise. Industry diversified mfg Dummy variable that takes a value of one if the buyer is active in the diversified manufacturing industry, and zero otherwise. Industry grocery Dummy variable that takes a value of one if the buyer is active in the groceries industry, and zero otherwise. Industry hard goods retail Dummy variable that takes a value of one if the buyer is active in the hard goods retail industry, and zero otherwise. Industry soft goods retail Dummy variable that takes a value of one if the buyer is active in the soft goods retail industry, and zero otherwise. Supplier large size Dummy variable that takes a value of one when the total sales of the supplier exceeds US$ 2 billion, and zero otherwise. Supplier medium size Dummy variable that takes a value of one when the total sales of the supplier is between US$ 100 million and US$ 2 billion, and zero otherwise. Supplier small size Dummy variable that takes a value of one when the total sales of the supplier is less than US$ 100 million, and zero otherwise. Supplier investment grade Dummy variable that takes a value of one when the supplier is investment grade, and zero otherwise. 32