Supply Chain Characteristics and Bank Lending Decisions

Transcription

1 Supply Chain Characteristics and Bank Lending Decisions Iftekhar Hasan Fordham University and Bank of Finland 45 Columbus Circle, 5 th floor New York, NY Phone: ihasan@fordham.edu Kristina Minnick Department of Finance Bentley University Adamian Academic Center Waltham, MA Kartik Raman Department of Finance Bentley University Adamian Academic Center Waltham, MA 02452

2 Supply Chain Characteristics and Bank Lending Decisions Abstract This paper investigates whether borrowers supply chain relationships affect banks lending decisions. These relationships benefit firms by reducing the information gap with banks, which increases the access to capital, while reducing the cost of the loan. However, banks demand increased intensity of covenants and greater use of collateral when the correlation of cash flows among firms in the supply chain is high. Longer relationships between the borrower and its supply chain partner, and between the bank and the borrower s supply chain partner, mitigate lending constraints. The evidence suggests supply chains serve as an informational bridge between lenders and borrowers.

3 A fundamental consideration in bank lending decisions is the borrowing firm s performance. Given that a borrower s supply chain constitutes an integral part of its operations, an important question arises: What factors determine a bank s decision to lend to multiple firms along a given supply chain? We examine this question empirically. Economic theories offer potential reasons why supply chain attributes should matter for bank lending decisions. Banks develop close relationships with borrowers over time, and consequently, they learn more about the borrower s characteristics than other banks can. 1 As a result, competing banks outside the lending relationship face relatively high information gathering costs and find it difficult to break into existing banking relationships. Supply chains provide an avenue for banks to use information from existing lending relationships to consider loans to other firms in the same chain. Correlated cash flows across firms in a shared supply chain enable a bank to verify information about those firms with greater certainty. By lending to a current borrower s supply chain partner, a bank may be able to better anticipate industry conditions and trends, as well as gain better insight into the partner s factors of production. This understanding can improve the bank s ability to monitor and narrow the information gap that outside lenders typically face. In addition, through repeated interactions with the borrower, the bank could acquire information about that firm s supply chain partners at a lower marginal cost than would be possible without an existing lending relationship (e.g., Petersen and Rajan, 1994; Bharath, Dahiya, Saunders, and Srinivasan, 2007). Close relationships between suppliers and customers are likely to reflect favorable private information by suppliers about their customers and thereby alleviate the bank s lending problem arising from asymmetric information (Biais and Gollier, 1997). The lower marginal cost of acquiring information about 1 See, e.g., Kane and Malkiel, 1965; Leland and Pyle, 1977; Diamond, 1984; Ramakrishnan and Thakor, 1984; Fama, 1985; Diamond, 1991; Rajan, 1992; Rajan and Winton, 1995; Mester, Nakamura, and Renault,

4 supply chain firms enhances the bank s ability to offer loans to existing borrowers supply chain partners; thereby improving the bank s potential to expand business. Alternately, the positive correlation among cash flows of firms in a supply chain leads to contagion effects when disruptions occur at one or more firms in the chain. The default risk for a bank lending to multiple firms in a given supply chain thus increases. Empirical evidence about the effects of financial distress on the firm s suppliers, reported in Hertzel, Li, Officer, and Rodgers (2008), supports this proposition. Portfolio theory suggests that closer ties between firms and their supply chain partners should reduce the likelihood that a bank would lend to both a firm and its supplier (and/or customers). We empirically examine a bank s decision to lend to multiple firms along a given supply chain using a large sample of customer-supplier relationships and bank loans covering more than 43,000 firm-years from 1992 through Our key findings are as follows. First, about 18% of supply chain relationships in our sample involve loans by the same bank to both the supplier (upstream party) and the customer (downstream party). Furthermore, when the same bank lends to both the supplier and customer, in more than 85 percent of cases, the bank lends to the supplier first before lending to the customer. Because the customers in our sample are substantially larger in size (and hence borrow larger amounts) than their corresponding suppliers, this pattern in the data suggests that banks potentially manage their exposure to losses on large loans by using information acquired from the smaller loan to the upstream party in order to make decisions about the larger loan to the downstream party. The results from multivariate regressions controlling for firm and supply chain partner characteristics indicate that the probability of a supplier receiving a loan from a given bank is 53% if the bank has an existing lending relationship with the corresponding customer, but only 2

5 17% without such a relationship, which is economically significant. We find a similar pattern for the likelihood of customers receiving a loan from a given bank, although the difference in probability is less pronounced. These findings suggest that firms benefit from supply chain relationships by reducing potential information gaps with banks, thereby improving their access to capital. Second, we identify supply chain characteristics that could potentially explain a bank s decision to lend to multiple firms along a given supply chain. When the supplier has a bank loan, the same bank is less likely to lend to the corresponding customer if the: (i) supplier depends more on the customer, (ii) supplier and customer belong to the same industry, (iii) supplier has a more concentrated customer base, or (iv) supplier s cash flow is more positively correlated with that of its customer. We find similar results for the bank s decision to lend to the supplier, conditional on the customer having a loan from the same bank. Given that the preceding variables reflect the extent to which the customer s and supplier s respective cash flows are closely integrated, the results imply that banks are less likely to lend to both upstream and downstream firms when the potential for contagion risk along the supply chain is higher. We also find when a supplier has a bank loan, the same bank is more likely to lend to the customer when the duration of the supplier s relationship with the customer is longer. This result suggests that an existing lending relationship with the supplier enables the bank to lower its marginal cost of gathering information about the customer, thereby enhancing the bank s ability to expand business by competing effectively for the loan to the customer. At the same time, the result also suggests that the supply chain relationship s stability (reflected by its length) mitigates the bank s perceived risk of lending to the downstream party, increasing the likelihood of the bank lending to the borrower s customer. 3

6 Third, supply chain characteristics influence the pricing of loans by a given bank to both the existing borrower and to that borrower s supply chain partners. Controlling for endogeneity arising from the bank s choice to lend, we find that the yield spread for the supplier is higher if: (i) the supplier is more dependent on a customer for its sales, (ii) the supplier operates in the same industry as its customer, and (iii) the correlation between the cash flows of the supplier and customer is more positive. Furthermore, the adverse effect on the yield spread resulting from a supplier s dependence on the customer is less pronounced if the supplier s relationship with its customer is of longer duration. These findings suggest that although correlated cash flows between supply chain partners increase the bank s perceived risk from contagion, the stability of customer-supplier relationships mitigates this risk. We also examine whether, in addition to the borrower s relationship with its supply chain partner, the bank s relationship with the supply chain partner matters for the pricing of the borrower s loan. The bank s existing relationship with a prospective borrower s supply chain partner is likely to influence the extent to which the bank acquires information about the borrower prior to the loan. Consistent with this intuition, we find that the borrower incurs a lower yield spread if the duration of the bank s relationship with the borrower s supply chain partner is at least five years. Given this informational benefit of lending to multiple firms along the supply chain, we further test the importance of the bank s relationship with the supply chain partner in the context of evidence reported in recent studies. Campello and Gao (2016) show that firms with high customer concentration are more likely to default on their bank loans, and consequently incur higher yield spreads. Like Campello and Gao, we find that high customer concentration (measured as the percentage of a supplier s sales attributable to a single customer) is indeed 4

7 associated with higher yield spreads. Importantly, in the context of our study, we show that the adverse effect of high customer concentration on loan spreads is mitigated when the bank s relationship with the borrower s customer is at least five years old. Furthermore, Cen, Dasgupta, Elkamhi, and Pungaliya (2016) find that firms with a longer duration of relationship with the principal customer incur lower yield spreads. We confirm this finding and show that the benefits of longer customer-supplier relationships are further enhanced through even lower yield spreads when the bank s relationship with the borrower s customer is at least five years old. These results indicate that banks incorporate information from their existing lending relationship with a prospective borrower s supply chain partner when deciding the terms of the loan for the borrower. More broadly, the results suggest that existing lending relationships serve as a conduit for banks to acquire information and thereby facilitate new loans to other firms along a given supply chain. We also find that the timing of loans matters for yield spreads. The effect of supply chain characteristics on a supplier s yield spread is more pronounced if the supplier receives the loan after the customer does. Given that contagion effects of default along the supply chain are more severe for suppliers than for customers, as shown by Hertzel, Li, Officer, and Rodgers (2008), this finding is consistent with the explanation that banks consider the effect of contagion risk in their decision to lend to multiple firms along a supply chain. In general, we find similar results in specifications explaining the customer s loan spreads. Unlike for suppliers, however, the results for customers are not sensitive to whether the customer received the loan first, consistent with the notion that customers are less susceptible to contagion risk along the supply chain. Furthermore, poorer credit ratings for the customer (supplier) are associated with higher loan 5

8 spreads for the supplier (customer), consistent with the idea that a poorer credit rating for the borrower s supply chain partner is more likely to add to the bank s risk from contagion. Lastly, controlling for the risk of the borrower, and for other economic determinants identified in the literature, firms are associated with a greater intensity of loan covenants and are more likely to secure the loan using collateral if: (a) they are more dependent on their supply chain partner, (b) their cash flows are more correlated with those of their supply chain partner, (c) the duration of their relationship with the supply chain partner is shorter, and (d) the borrower s supply chain partner has a poor credit rating. Overall, the results suggest that banks incorporate information about the borrower s supply chain in setting the number of loan covenants and in deciding whether to require collateral to secure the loan. Given that the terms of bank loans are determined endogenously, and given the potential omission of unobserved variables in our empirical specifications, we re-examine the specifications explaining the price and non-price terms of bank loans using the geographic distance between the firm and its supply chain partner as an instrument for variables that measure the strength of the supply chain relationship. We find that the conclusions are unchanged based on the instrumental variable specifications. To our knowledge, this study is one of the first to examine the decision by a bank to lend to multiple firms along a given supply chain. Our paper joins a growing stream of research on the importance of supply chain relationships to the terms of bank loan contracts (Cen, Dasgupta, Elkamhi, and Pungaliya, 2016; and Campello and Gao, 2016). Prior studies also show that closer ties between borrower and lender improve the availability and terms of funds for the borrower (see Berlin and Mester, 1992, 1998; Boot, Greenbaum, and Thakor, 1993; Petersen and Rajan, 1994; and Bharath, Dahiya, Saunders, and Srinivasan, 2011). Our study contributes to and 6

9 extends this literature by presenting evidence that lending relationships could also benefit the existing borrower s supply chain partners. Moreover, from a lender s perspective, Bharath, Dahiya, Saunders, and Srinivasan (2007) find that relationship lending significantly enhances a bank s likelihood of providing a future loan to an existing borrower, thereby making it difficult for a non-relationship lender to compete for the same loan. Our findings extend this line of research by suggesting that supply chain relationships enhance banks ability to lend to existing borrowers supply chain partners, thereby overcoming the difficulty of gaining new business beyond existing clients. From a borrower s perspective, our study extends the idea that strong banking relationships influence a firm s ability to raise additional capital (e.g., Rajan, 1992; Gopalan, Udell, and Yerramilli, 2011). Our findings suggest that firms can mitigate the costs associated with information asymmetry and potentially enhance their access to credit by seeking capital from their supply chain partner s banks. Another strand of the literature examines the consequences of physical distance between lenders and borrowers for the availability and terms of loans (Petersen and Rajan, 2002; Kim, Kliger, and Vale, 2003; Degryse and Ongena, 2005). In particular, Petersen and Rajan (2002) find that technological improvements mitigate the adverse effects of distance on lending activities by making it easier for banks to obtain and verify hard information about potential borrowers. Although our study does not focus on physical distance, our findings suggest that supply chains offer an informational channel between a lender and the supply chain partners of an existing borrower, through which information about prospective borrowers flows to the lender. This connection enhances the lender s likelihood of expanding business beyond the existing borrower. 7

10 Our study also contributes to research examining the effects of firms product market relationships on the cost of bank loans. For instance, Valta (2012) finds that intense rivalries in the product market induce greater default risk and lower liquidation values for borrowers, leading to higher costs of bank financing. In contrast, we focus on the borrower s supply chain relationships, and our results (controlling for industry fixed effects), suggest that borrowers with greater reliance on supply chain partners incur higher costs of bank loans due to the risk of contagion from default along the supply chain. In a broader context, our study contributes to research suggesting that non-financial stakeholders invest less in relationship-specific assets if high leverage increases the hold-up costs associated with supply chain disruptions (Kale and Shahrur, 2007; Banerjee, Dasgupta, and Kim, 2008). Consistent with this reasoning, we find that financial stakeholders, such as banks, also invest less in firms that are potentially exposed to risks arising from supply chain disruption. The paper proceeds as follows. We describe the data and sample characteristics in the next section. Section 3 presents a discussion of the empirical research design and results. Section 4 concludes. 2. Data and sample description To build the customer-supplier relationship database, we start with all firms covered by the Compustat Customer file in the Compustat Segment Database from 1992 through Identifying customer-supplier pairs requires matching abbreviated names of customers in the Customer file with the full company names on Compustat. Because we do the matching by hand (and the last year for which we have complete DealScan data is 2009), our sample period ends in In accordance with FASB Accounting Standards Codification, ASC , public firms may disclose the identity of any customer whose purchases represent more than 10 percent 8

11 of a firm s total revenues, although they are not mandated to do so. The procedure we follow to identify the customer firms, described in detail in the appendix, is similar to that used in Fee and Thomas (2004). To assemble the necessary data, we first identify 7,987 unique suppliercustomer relationships and 21,562 firm-year observations between 1992 and 2008 in which both the supplier and customer have a global Company Key (GVKEY) identifier and are listed in Compustat. We then obtain loan data from the Thomson Reuters LPC DealScan database. The basic unit of loans is a lending facility. A firm can obtain multiple facilities with the same loan package in a contract year, and loan terms can differ across these facilities. Consequently, we treat each facility-year as a distinct observation. We match the same firm-year information from Compustat to multiple facility-year observations if a firm obtains multiple facilities in a year. Using the DealScan-Compustat linked file provided by Michael Roberts (Chava and Roberts, 2008), we merge the bank loan sample with Compustat. Panel A of Table 1 presents the distribution of customer-supplier relationships categorized by whether or not a customer-supplier pair received at least one loan from the same bank during the year. After removing observations with incomplete DealScan or Compustat information, we obtain a final sample in which, out of 43,124 total supply-chain-year relationships, 5,094 suppliers and 12,441 customers received a loan. We next identify cases in which the same bank made a loan to both the customer and supplier. DealScan includes a number of descriptions. To ensure we do not mislabel a lead bank we follow a simple rule. Any bank(s) not described as a participant is treated as a lead bank, similar to Bharath, Dahiya, Saunders, and Srinivasan (2011). If a bank classified as a lead bank makes a loan to both the supplier and the customer during the supply chain relationship, and within t-3 to t+3 years, then we record that both supply 9

12 chain partners received a loan. 2 We find that 29% of the suppliers received a loan from a bank that already had an existing relationship with their customer, while 14% of the customers received a loan from a bank that had an existing relationship with the supplier. For example, Toys-R-Us is a customer of Hasbro. In 2005, Hasbro received a loan, and in 2006, Toys-R-Us received a loan from the same bank. Panel B of Table 1 presents the data at the bank facility level, wherein each loan to a supplier (or customer) represents a unique observation. The results in Panel B of Table 1 reveal that conditional on both the supplier and customer obtaining a loan from the same bank, banks first lend to the supplier in a majority of the cases (5,493 out of 6,136 supplier loans, or about 89%; and 6,029 out of 6,952 customer loans, or about 86%). Given that the average size of loans to customers is larger than that for the suppliers, this pattern in loan distribution suggests that banks appear to minimize their risk exposure while acquiring information about the supply chain by first lending to the smaller upstream firm before lending to the larger downstream firm. To summarize the results in Table 1, we find that an economically significant proportion of customer-supplier relationships involve loans by the same bank to both supply chain partners. We also find that banks prefer to lend to the smaller of the two parties first, suggesting that banks appear to minimize the risk to their loan portfolio by using the smaller loan to acquire information relevant for the larger loan. Table 2 shows the univariate statistics for both the relationship characteristics as well as the firm characteristics (definitions are provided in the appendix). Not surprisingly, given the nature of the database, we find that suppliers are more dependent on their customers than customers are dependent on their suppliers. The ratio of sales to the customer adjusted by total 2 We also use other cut-off dates that are more stringent (t-1 to t+1), and more lax (t-5 to t+5), with qualitatively similar results. 10

13 supplier sales is significantly larger than the ratio of sales to the customer over the customer s cost of goods sold (COGS). Most of the supply chain relationships appear to be horizontal (as opposed to vertical) in terms of industry membership, only 8% of the relationships are in the same industry. On average, supply chain relationships last almost four years, although 25% of them have lasted 11 years or longer. We also measure how synchronized the supply chain partners cash flows are by regressing the supplier s cash flow on the customer s cash flow over the prior 12 quarters. Synchronicity captures the R 2 of the estimations. We find that on average, 58% of the supplier s cash flows are explainable by their customer s cash flows. Consistent with the literature, the customers in our sample are substantially larger than the suppliers are, and have higher ROA and sales growth. The suppliers have higher cash ratios, leverage, and riskiness of earnings. Given the significant differences between customers and suppliers, we control for their respective characteristics in all our regression specifications. 3. Empirical design and results Our objective in this study is to explain the likelihood and terms of bank loans to multiple firms operating in the same supply chain. First, we focus on the likelihood that a firm receives a loan from a given bank, given that its supply chain partner has a loan from the same bank. Next, we focus on the sub-sample of all firms that received loans and examine whether an existing lending relationship with the firm s supply chain partner influences the likelihood that the firm also receives a loan from that same bank. We then examine how the supply chain relationship influences loan terms, focusing on loans in which both the customer and supplier have borrowed from the same bank Likelihood of a firm receiving a bank loan 11

14 As we discuss earlier, existing theories of relationship lending predict that strong relationships lower the marginal costs of information acquisition. A testable implication of this idea is that a lender is more likely to secure a prospective borrower s lending business if it has an existing relationship with the prospective borrower s supply chain partner. Using a logistic regression framework where the dependent variable equals one if the company receives a loan and zero otherwise, we estimate the following equation: GOTLoan i,j,t = β 1 + β 2 SCP Got Loan i,j,t 1 + β 3 8 Relationship Characteristics i,j,t 1 + β 9 14 Firm Characteristics i,j,t 1 + β SCPFirm Characteristics i,j,t 1 + ε, (1) where, i indexes firm, j indexes bank, and t indexes year. SCP Got Loan is an indicator variable that equals one if the supply chain partner has an existing loan from the bank. Relationship characteristics include CSALE Ratio, the size of the firm, the size of the supply chain partner, a Same Industry indicator, Relationship Length, and Synchronicity. Firm characteristics include MK to BK, Leverage, ROA, Cash Holdings, Sales Growth, and Earnings Volatility. SCP Firm Characteristics are the same as Firm Characteristics, but for the firm s supply chain partner. The results, reported in Panel A of Table 3, indicate a positive and significant coefficient on SCP Got Loan. This result indicates that the likelihood of receiving a loan from a given bank is higher for a firm whose supply chain partner has an existing loan from the same bank. The marginal effects, shown in Panel B, suggest that, conditional on the supply chain partner not having a loan versus having a loan from the same bank, suppliers experience a more pronounced increase in the probability of receiving a loan (from 17% to 53%). Considering only 23% of the suppliers on average have loans, this increase in likelihood that arises from having a supply chain partner with a loan from the same bank is economically significant. Customers see an increased probability of receiving a loan if their supply chain partners have a loan from the same bank 12

15 (from 56% if SCP does not have loan to 84% if SCP has a loan). On average, 57% of the customers have loans and so an increase 84% is economically significant. The results in Panel A of Table 3 also indicate that firms are more likely to receive a bank loan if their supply chain partner is smaller in size, and if the duration of their relationship with that partner is longer. In addition, suppliers are less likely to receive a loan if the supply chain partner belongs to the same industry as the supplier. Furthermore, the likelihood of a customer receiving a loan decreases if its cash flow is more positively correlated with that of its supplier. These results are generally consistent with banks attempting to minimize the risks associated with lending to multiple firms along the supply chain, including risks from supply chain disruptions as well as common industry shocks affecting all firms along the supply chain. The results for the control variables (Firm Characteristics) are as expected based on the existing literature. Chava, Livdan, and Purnanandam (2009) suggest that banks are less likely to lend to firms with higher leverage and if they do lend to them, incorporate that increased risk into higher spreads. We find similar results. Companies with higher leverage are less likely to receive a bank loan. We also find that firms with higher ROA and higher levels of cash holding are more likely to receive loans, consistent with the literature. We find that with respect to a firm s supply chain partner characteristics, firms are more likely to receive a bank loan if their supply chain partner is associated with higher growth and lower earnings volatility Role of supply chain characteristics in the bank s decision to lend to multiple firms along the supply chain In this section, we examine how supply chain characteristics influence the likelihood that a given bank lends to both the supplier and the customer. While lending to multiple firms along the supply chain entails risks for the bank arising from potential supply chain disruptions, the benefits of lending to both the supplier and customer include lower marginal costs of information 13

16 gathering and monitoring resulting from economies of scale. Empirically, we examine the likelihood that a firm receives a loan from a bank conditional on the bank having an existing lending relationship with the firm s supply chain partner. The specification of the logit estimation is as follows: BOTHGOTLoan t = β 1 + β 2 7 Relationship Characteristics t 1 + β 8 13 Firm Characteristics t 1 + β SCPFirm Characteristics t 1 + ε, (2) We conduct this test separately for suppliers and customers. Specifically, the dependent variable in columns 1 and 2 of Table 4 is defined as follows: Conditional on the supplier having a bank loan during the year, the dependent variable equals one if the customer also has a loan from the same bank during the same year, and equals zero otherwise. Similarly, in columns 3 and 4, conditional on the customer having a loan, the dependent variable equals one if the supplier also has a loan from the same bank. The results in columns 1 and 2 of Table 4 indicate a negative and significant coefficient on CSALE Ratio (defined in columns 1 and 2 as the fraction of the supplier s total sales revenue attributable to the given customer). Thus, conditional on the supplier having a bank loan, banks are less likely to lend to the customer if the supplier is more dependent on the customer for a large fraction of its sales revenue. This finding is consistent with banks considering the borrower s supply chain characteristics in minimizing the risks to their loan portfolios. If a bank lends to the customer, it is not only exposed to the risk that the customer might perform poorly, but also to the contagion effect on the supplier s performance, thereby putting the bank s existing loan to the supplier at risk. Thus, the result suggests that banks minimize risk by avoiding lending to a customer if the supplier depends heavily on that customer. 14

17 Furthermore, we find a significant and negative coefficient on Same Industry and Synchronicity, which indicates that banks are reluctant to lend to the customer if the supplier and customer are exposed to common risk factors and if their cash flows are highly correlated. By indicating the extent to which the cash flows of the customer and supplier are likely to be closely integrated, these explanatory variables reflect the potential for contagion risk along the supply chain arising from default by one of the parties in the chain. Thus, the results in columns 1 and 2 suggest that banks are less likely to lend to both upstream and downstream firms if the potential for contagion risk along the supply chain is high. Controlling for these factors, we also find a positive and significant coefficient on Relationship Length in columns 1 and 2, indicating that conditional on the supplier having a loan from a given bank, a bank is more likely to lend to the customer when the duration of the supplier s relationship with the customer is longer. This result is consistent with two nonmutually-exclusive explanations. Given the bank s existing lending relationship with the supplier, a longer relationship between the customer and supplier enables the bank to lower its marginal cost of gathering information about the customer, thereby enhancing the bank s ability to compete for loans to the customer. At the same time, a longer relationship between the supplier and customer implies a more stable supply chain, which mitigates the bank s perceived risk of contagion from default along the supply chain, thereby increasing the likelihood of the bank lending to the customer. The results in columns 3 and 4 are largely similar to those reported in columns 1 and 2. The interpretation of the results in columns 3 and 4, however, is from the perspective of a supplier receiving a loan from the bank, conditional on the customer having an existing loan from the same bank. For example, the negative coefficient on CSALE Ratio indicates that banks 15

18 are less likely to lend to the supplier if the customer is more dependent on the supplier for its cost of goods sold. Notably, the coefficient of Relationship Length is insignificant (albeit positive) in columns 3 and 4, indicating that the duration of the customer s relationship with the supplier does not significantly affect the likelihood of the bank lending to the supplier. The coefficients on control variables in columns 1 and 2 of Table 4 indicate that conditional on the supplier having a loan from a given bank, some characteristics of the supplier matter in determining the likelihood of the customer receiving a loan from the same bank. In particular, a customer is more likely to receive a loan from the same bank if the supplier has more valuable growth opportunities (higher MK to BK), lower leverage, and higher profitability (higher ROA). These results suggest that banks evaluate the future growth prospects and default risk potential of upstream firms in deciding whether to lend to downstream firms. Furthermore, consistent with expectations, we find that banks are more likely to lend to the customer if the customer has higher cash holdings and less volatile earnings. In Panel B of Table 4, we present the predicted probabilities corresponding to the regression coefficients in Panel A, moving from the 25 th percentile to the 75 th percentile values of a given variable, holding the other variables at their respective means. The results in Panel B indicate that conditional on one of the supply chain partners having a bank loan, the probability of the other partner receiving a loan from the same bank declines by 6 percent if the dependence of one partner on the other increases from the 25 th percentile to the 75 th percentile. Given that on average only 28% of the suppliers have loans if their partners also have a loan, a decrease of 6% is economically significant. Similarly, the probability of receiving a loan from a given bank declines between 3 percent and 5 percent if both supply chain partners belong to the same industry rather than different industries. These illustrative findings suggest that banks pay 16

19 attention to supply chain characteristics in making lending decisions. Overall, the results in Table 4 suggest that banks consider the degree of inter-dependence between an existing borrower and its supply chain partners in determining whether to lend to the supply chain partners Pricing of bank loans Next, we examine whether supply chain characteristics affect the yield spreads (over LIBOR) for borrowing firms. Because we are interested in understanding the effects on yield spreads arising from bank loans to multiple firms along the supply chain, for this analysis, we only include observations where both the customer and the supplier have a loan from the same bank. Table 5 shows the univariate characteristics of the loan facilities and bank characteristics, segmented by suppliers and customers. We find that suppliers incur significantly higher yield spreads, have worse ratings, require more covenants, and are more likely to have secured loans compared with customers. We also find that for a vast majority of loans, the borrower s supply chain partner has a recent loan (less than one year old) from the same bank, whereas in about five to nine percent of the loans, the supply chain partner has a relatively longstanding loan (at least five years old) from the same bank. There is no significant difference between the size, deposit ratio, and capital ratio for banks that lend to the suppliers and customers. Previous studies identify borrower characteristics affecting the yield on bank loans (see, e.g., Berger and Udell, 1995; Graham, Li, and Qiu, 2008; Bharath, Dahiya, Saunders, and Srinivasan, 2011; Hasan, Hoi, Wu and Zhang, 2014). We control for previously identified determinants of the yield spread on bank loans and examine whether the characteristics of the borrower s supply chain affect yield spreads for the borrowing firm. We also include as a control variable the Inverse Mills Ratio (IMR) from the logistic regression predicting the likelihood of the supplier or customer receiving a loan from the bank conditional on the supply chain partner having a loan 17

20 from the same bank. The specification to estimate the IMR for Model 1 and 2 in Table 6 (Supplier loans) is based on Model 2 from Panel A of Table 4, and the IMR for Model 3 and 4 in Table 6 (Customer loans) is based on Model 4 in Panel A of Table 4. We conduct the analysis, with supplier loan spreads as the dependent variable, on two separate sub-samples: (i) supplier receives the loan first, i.e., before the customer receives a loan from the same bank, and (ii) supplier receives the loan after the customer receives a loan from the same bank. In the latter sub-sample, we expect a more pronounced relationship between risk factors and loan spreads because of the added risk to the bank s portfolio from exposure to default by both, the customer and the supplier. We apply the same logic and examine the loan spreads for customers. Focusing on the sub-sample where both a firm and its supply chain partner receive a loan from the same bank, we estimate the following specification: Spread t = β 1 + β 2 IMR + β 3 11 Relationship Characteristics t 1 + β Firm Characteristics t 1 + β Bank Loan Characteristics t 1 + β Bank Characteristics t 1 + ε, (3) The results in column 1 of Table 6, with supplier loan spread as the dependent variable, indicate a positive and significant coefficient on CSALE Ratio and Same Industry. If the supplier s partner is in the same industry, the bank will increase the spread on the loan by 13% to compensate for the additional risk. In the sub-sample where the supplier receives the loan after the customer (column 2 of Table 6), we find a positive and significant coefficient on Synchronicity, which suggests that banks will increase the loan spreads by 14% due to the increased risk. Furthermore, in column 2, the negative and significant coefficient on Loan Supplier/Loan Cust indicates that the bank assesses a higher yield spread on the supplier s loan when the size of the customer s existing loan with the same bank is larger. This finding suggests 18

21 that banks incorporate information about supply chain links in anticipating the risks to their overall loan portfolios. These results suggest that banks price contagion risk in the supply chain by imposing higher borrowing costs on suppliers: (i) that are more dependent on customers, (ii) whose cash flows are likely to be more positively correlated with those of their key customers, and (iii) whose customers have larger existing loans with the same bank. The results in column 2 of Table 6 also indicate that if the supplier receives a loan from a given bank after the customer has obtained a loan from the same bank, the supplier s yield spread is lower if the duration of its relationship with the customer is longer (based on the negative coefficient on SCP Relat Length). A one standard deviation increase in the length of the supply chain relationship reduces the spread by 4%. Similarly, if the length of the bank s existing lending relationship is five years or greater, the spread is 12.3% lower (based on the negative coefficient on Old Bank SCP Rel). Lending relationships enable banks to gather information about a borrower s supply chain over time. We thus expect a bank to face lower marginal costs of acquiring information about the supplier s loan as the bank is likely to have gathered some of the information through its existing lending relationship with the customer. As a consequence of the lower marginal costs incurred by the bank, we suggest that the bank is able to offer a lower rate to the supplier. The results also suggest that the stability of relationships between the borrower and its supply chain partner, as well as between the bank and the supply chain partner, mitigates the bank s perceived risk of supply chain disruption, thereby lowering the cost of borrowing. We also find a positive and significant coefficient on both Rating and Rating SCP in all the models. We note that higher values of Rating imply worse credit quality. This result indicates that firms incur higher borrowing costs if their supply chain partners have poorer credit ratings, 19

22 providing further evidence that banks evaluate the characteristics of the borrower s supply chain partners and determine loan yields by incorporating the risk of contagion in the supply chain. From the borrower s perspective, borrowing costs are lower for firms that have supply chain partners with good credit ratings. The coefficients on control variables in Table 6 are largely consistent with the literature and with expectations. We find that the yield spreads are generally lower for syndicated loans, for loans that are secured and for loans that have a higher intensity of covenants. Yield spreads are also lower for less levered firms, for firms that are more profitable (based on ROA), and for firms with less volatile earnings Interaction effects In this section, we examine whether the bank s ability to acquire information through the supply chain mitigates the lending risk, and thereby lowers the yield spread. Consistent with Campello and Gao (2016) - where borrowers with a high degree of customer concentration for found to incur higher yield spreads - we find a positive coefficient on CSALE Ratio, indicating that suppliers incur higher yield spreads when a larger proportion of their sales is attributable to a principal customer. More importantly for our study, we interact CSALE Ratio with Old Bank SCP Rel, an indicator variable that equals one if the bank has an existing relationship with the borrower s customer. In Table 7, the regression result, with yield spread as the dependent variable, shows a negative coefficient on the interaction term. This result indicates that the adverse effect of higher supply chain dependence (higher CSALE Ratio) on the yield spread is mitigated by a longer duration of relationship between the bank and the borrower s supply chain partner. Looking at the standardized coefficients, a one standard deviation increase in the value of the interaction reduces the loan spread by 11% for suppliers and by 3% for customers. 20

23 Additionally, consistent with Cen, Dasgupta, Elkamhi, and Pungaliya (2016) where the yield spread is lower for borrowers having a longer duration of relationship with their principal customer - we report that the coefficient of SCP Relat Length is indeed negative and significant. Interacting CSALE Ratio with SCP Relat Length, we find that a one standard deviation increase will reduce suppliers loan spreads by 13% and customers loan spreads by 8.1%. This result combines the evidence in Campello and Gao (2016) and that in Cen, Dasgupta, Elkamhi, and Pungaliva (2016) by suggesting that the adverse effect of having high customer concentration is mitigated for the borrower when the borrower has a longer relationship with its principal customer. In order to relate our findings to those of Cen, et al., we interact SCP Relat Length with Old Bank SCP Rel. This interaction captures the significance of having strong supply chain relationships and strong lending relationships. The result in Table 7 shows a negative and significant coefficient on the interaction term, indicating that the yield spread is further reduced when the borrower has a long relationship with the principal customer and when the bank has a long lending relationship with the borrower s supply chain partner. A one standard deviation increase in the interaction reduces the spread by 12% for suppliers and by 4.8% for customers. Collectively, the results in Table 7 suggest that the supply chain information acquired by banks over time mitigates their lending risk and thereby enables them to lend to other firms along the given supply chain at a reduced yield Supply chain characteristics and the intensity of loan covenants and collateral Prior research suggests that riskier firms are associated with a larger number (i.e., greater intensity) of covenants and a higher likelihood of collateralization in their loan contracts (e.g., Demiroglu and James, 2010). The intuition is that greater intensity of covenants and/or the 21

24 requirement of collateral provide multiple avenues for banks to exercise the option to intervene in case the borrower displays signs of financial distress (see e.g., Berlin and Mester, 1992). We examine the role of covenants and collateralization in mitigating the risks faced by banks that lend to multiple firms along a given supply chain using the following two models: Log(Covenant) t = β 1 + β 2 IMR + β 3 11 Relationship Characteristics t 1 + β Firm Characteristics t 1 + β Bank Loan Characteristics t 1 + β Bank Characteristics t 1 + ε, (4) Secured t = β 1 + β 2 IMR + β 3 11 Relationship Characteristics t 1 + β Firm Characteristics t 1 + β Bank Loan Characteristics t 1 + β Bank Characteristics t 1 + ε, (5) The dependent variable in columns 1 through 4 in Table 8 is loan covenant intensity, measured as the log of the total number of covenants in a loan facility. In this analysis, we control for previously identified determinants of loan covenant intensity, and we include the IMR from the logistic regression predicting the likelihood of a supplier or customer receiving a loan from the bank conditional on the supply chain partner having a loan from the same bank. The specification used to estimate the IMR for Model 1, 2, 5, and 6 in Table 8 (supplier loans) is based on Model 2 from Table 4, and the IMR for Model 3, 4, 7, and 8 in Table 8 (customer loans) is based on Model 4 in Table 4. The results in Table 8, with covenant intensity as the dependent variable, indicate a positive and significant coefficient on variables that proxy for higher contagion risk borne by the lender resulting from default by one of the parties in the supply chain. In general, we find a positive and significant coefficient on CSALE Ratio, Same Industry, and Synchronicity. For instance, the marginal effects of the poison model shows a one unit increase in CSALE Ratio 22

25 increase the covenant intensity between % for suppliers, and 5-19% for customers. These results suggest that banks mitigate the risks arising from supply chain disruptions by imposing a higher intensity of covenants, thereby enhancing their ability to intervene if the borrower s probability of default increases. We also find a negative and significant coefficient on Old Bank SCP Rel in both Models 2 and 3, and the marginal effects show that firms where their partners have a long-standing relationship with the bank have between 1% -11% reduction in the intensity of their covenants. We also find a negative and significant coefficient on Recent Bank SCP Rel in Model 3. These results suggest that the strength of the relationship between the lending bank and the borrower s supply chain partner mitigates the bank s perceived risk thereby reducing the intensity of covenants. A potential explanation for this result is that the bank acquires information about the borrower through the existing lending relationship between the bank and the borrower s supply chain partner, which reduces the information asymmetry faced by the bank when lending to the borrowing firm. We also find a negative and significant coefficient on SCP Relat Length, which indicates that the intensity of loan covenants is lower if the duration of the borrower s relationship with the supply chain partner is longer. This finding is consistent with the explanation that banks perceive a lower risk of supply chain disruption or of default if the borrower has a more stable and longstanding relationship with the supply chain partner. Another important result in Table 8 is the positive and significant coefficient on Rating SCP. This result indicates that if we control for the borrower s own credit rating, banks perceive a higher default risk when the borrower s supply chain partner has a poorer credit rating. This evidence is consistent with the results reported earlier for loan spreads, and more importantly, 23

26 suggests that banks incorporate the risk of contagion resulting from default by one or more of the borrower s supply chain partners in determining the intensity of covenants in loan contracts. Columns 5 through 8 in Table 8 estimate Equation (5), in which the dependent variable is equal to one if the loan is secured and zero otherwise in the logit estimation, controlling for industry and year fixed effects and clustering standard errors at the firm level. Based on the coefficient of CSALE ratio, we find that the greater a firm s dependence on its supply chain partner, the higher the likelihood that the firm has to secure the loan (the marginal effects show that a one unit increase in CSALE ratio increases the likelihood of securing the loan between 6-145% for suppliers and 5-18% for customers. This finding may result from increased risk of potential supply chain disruption arising from the dependence. Moreover, firms in the same industry as their supply chain partner are more likely to secure the loan. Moreover, the more stable and longer the relationship between the firm and its supply chain partner, the smaller the likelihood that the firm secures the loan. Finally, the probability of the loan being secured is smaller if the supply chain partner has either a recent or a long-standing relationship with the bank. The coefficients of control variables are in line with expectations, and generally indicate that riskier firms are associated with a higher likelihood of having secured loans. Consistent with the findings in Demiroglu and James (2010), our results indicate that firms that are smaller, less profitable, more highly levered, with poorer credit ratings are associated with more intense loan covenants. Overall, the results in Table 8 support the notion that banks consider the risks emanating from supply chain partners in determining the intensity of covenants and collateral requirements in loan contracts Robustness checks: Endogeneity 24