The signaling role of trade credit in bank lending decisions: Evidence from small and medium-sized enterprises

1 INTRODUCTION

Trade credit is a widespread prominent source of financing for most firms and plays a crucial role in smaller firms. Empirical regularities refer to trade credit as the second most important source of external debt financing for small and medium-sized enterprises (SMEs; e.g., Demirguc-Kunt & Maksimovic, 2001).

The relevance of trade credit to SME financing led Berger and Udell (2006) to include trade credit among lending technologies, even if not provided by financial intermediaries. In the firm financing paradigm, trade credit is also considered a safety valve, especially for firms experiencing funding constraints (Molina & Preve, 2012; Petersen & Rajan, 1994). Studies on the recent financial crisis assert the role of trade credit in liquidity provision and the last resort lender, documenting a high trade credit dependence for constrained firms (Carbò-Valverde et al., 2016; Garcia-Appendini & Montoriol Garriga, 2013). Despite the relevance of the topic, empirical evidence on the ability of trade credit to reduce asymmetric information in lending decisions (e.g., loan approval) remains scarce. In general, information asymmetries require expensive screening and monitoring activities (Diamond, 1984). However, the effects can be mitigated, resulting in more favorable conditions in loan agreements (Chang et al., 2020; Cole, 1998; Cole et al., 2004; DeFranco et al., 2017; Ivashina 2009). Several studies show how the collection of private information should ameliorate asymmetric information issues in lending decisions. Among them, Biais and Gollier (1997) consider trade credit to be a signal that reveals private information to lenders. They introduce a theoretical signaling model that assumes that suppliers have an informational advantage in screening borrowers. Because suppliers have private information about their customers, they should be more effective in monitoring (Miwa & Ramseyer, 2008) and achieving cost savings in information collection (Jain, 2001). A wide range of empirical studies exploits the relationship between trade and bank financing to test the Meltzer hypothesis of the substitution effect among two forms of firm financing (Meltzer, 1960). Studies investigating the role of trade credit in lending decisions have provided limited evidence (e.g., Cook, 1999), and some studies are hampered by a lack of data related to bank lending decisions because they infer results based on the total debt of firms rather than internal information related to banks’ loan approval process (e.g., Agostino & Trivieri, 2014; Alphonse et al., 2004).

Based on this perspective, this paper aims to fill this gap by empirically examining the role of trade credit in terms of account payables in the loan approval process using inside information on the bank–firm relationship. Specifically, this study attempts to assess the role of trade credit in bank lending decisions (i.e., approving or denying a loan request) rather than focusing only on loans already granted as is the case in all previous studies. Therefore, we provide evidence of a complementary relationship between trade credit and bank loans based on a new perspective (the role of trade credit in the loan approval process) that has never been previously explored. In particular, we find that trade credit increases the probability of a borrower receiving a positive response to a loan request. This result may be related to a reputational mechanism based on which the trade credit provided by suppliers can be interpreted by a bank as a signal of the buyer's good reputation in the market. Our results show that this reputational mechanism is beneficial for all SMEs, especially those characterized by a higher level of opacity, which, given this specific aspect, may be subject to a credit rationing problem more than other SMEs (Berger & Udell, 1998).

These results contribute to the growing literature on the effect of credit information sharing on opaque firms, particularly in the SME context. The evidence also corroborates the importance of relationships in small business lending because relationships are still relevant and can benefit borrowers, especially those characterized by high intensity (Cole & Sokolyk, 2016). However, additional evidence on corporate failure extracted from our data does not exclude the existence of false signals related to banks’ lower monitoring ability with respect to trade creditors in the assessment of prospective debtors’ creditworthiness and ongoing activities as suggested by Biais and Gollier (1997).

This paper extends the literature in several ways. First, by exploiting unique data on bank lending rationing decisions, we directly address and explicitly test the signaling role of suppliers in lending decisions by using actual bank–firm data. Furthermore, we demonstrate that the substitution and complementary effects of trade credit on banks are not alternative effects such that a firm may prefer to use trade credit as the main form of short-term financing. Simultaneously, trade credit may be considered useful information for banks when making decisions regarding a loan request by the same firm.

The remainder of this paper is organized as follows. Section 2 reviews the related literature, and Section 3 develops the research hypotheses. Section 4 describes the data and provides summary statistics. Sections 5 and 6 report the methodology and discuss the results. Section 7 contains additional evidence on the role of trade credit in corporate failure, and Section 8 concludes.

2 RELATED LITERATURE

The literature on crises and their aftermath confirms that credit rationing to the SME sector has been economically significant (e.g., Ferrando et al., 2017; Presbitero et al., 2014). Previous empirical papers establish that financially constrained firms use trade credit to a larger extent (e.g., Demirguc-Kunt & Maksimovic, 2001; Fisman & Love, 2003; Petersen & Rajan, 1997). Recent studies jointly examine the effect of a crisis in terms of credit rationing and alternative forms of financing for small firms, validating the role of trade credit as a safety valve (Casey & O'Toole, 2014). Specifically, trade credit plays a crucial role in the type of lender of last resort during financial shocks (Taketa & Udell, 2007), conditioning the investment decisions of constrained firms in terms of capital expenditures (Carbò Valverde et al., 2016).

A plethora of studies have investigated trade credit and the motivations behind selling or buying products and services on credit (e.g., Elliehausen & Wolken, 1993). The common research question is why non-financial firms are involved in financial intermediation. This research can be split into two main literature strands concerning the motives behind trade credit usage that arise on the demand and supply sides. From the demand side, control protection and pooled transactions, which reduce the uncertainty of flow, reduce costs and constitute valid motives for selling products on credit (Chant & Walker, 1988; Ferris, 1981; Honoworth & Reber, 2003; Ono, 2001; Summers & Wilson, 2002; Wilner, 2000). Conversely, capital access, the informational advantage of goods suppliers, price elasticity, the easy repossession of collateral value and buyers’ earning quality are the principal reasons for the use of trade credit on the supply side (Biais & Gollier, 1997; Brennan et al., 1988; Danielson & Scott, 2004; Deloof & Jegers, 1999; Duarte et al., 2016; Emery, 1984; García-Teruel et al., 2014; Giannetti et al., 2011; Jain 2001; Petersen & Rajan, 1997; Pike & Cheng, 2001; Schwartz, 1974; Van Horen, 2007). Most of these theoretical arguments are also supported by empirical evidence. For instance, the findings of Petersen and Rajan (1997) show that trade credit usage is motivated by price discrimination. Additionally, Long et al. (1993) empirically validate the use of trade credit as a guarantee of product quality. The empirical findings also confirm that capital access, in terms of bank credit availability, affects the supply of trade credit. In particular, Petersen and Rajan (1994) find that relying less on trade credit results in a lower likelihood of being bank constrained. From this perspective, recent empirical studies have not produced unique results. Fabbri and Klapper (2008), who test the market power hypothesis, also test the importance of financial constraints. Their study reveals that access to bank financing and profitability are not significantly related to trade credit supply in the Chinese market. Love and Zaidi (2010) offer other evidence on Asian countries.

The findings of the authors above do not support the hypothesis that trade credit can help constrained SMEs during a crisis because financially constrained firms reduce their use and extension of trade credit. In contrast to the results of Tsurata (2015), in Japan, small businesses increase trade credit (both payables and receivables) if bank loan availability improves. In Europe, the findings confirm that credit-constrained firms are more likely to use and apply for trade credit (Casey & O'Toole, 2014); the same holds in the United States (Danielson & Scott, 2004) in the presence or absence of a shock. One possible explanation that can reconcile these findings is offered by Taketa and Udell (2007), who conclude that the relationship between trade credit and financial institution channels (complements or substitutes) can change across credit crunch periods and bubble periods. The effect of a shock depends on its typology. In particular, a negative (e.g., a credit crunch) and a positive (e.g., a bubble) financial shock can have different effects on trade credit and financial institution channels. Additionally, evidence from the extant empirical literature suggests that a nation's lending infrastructure affects financing decisions, which is in line with the theoretical framework of Berger and Udell (2006). Trade credit is particularly important in countries in which the financial system is not well developed, and industries with a higher dependence on trade credit exhibit higher growth rates (Demirguc- Kunt & Macksimovic 2001; Fisman & Love, 2003).

Despite the large number of empirical studies on trade credit motives collected in the review by Seifert et al. (2013), there is very little evidence on the informational competitive advantage of suppliers. The first assumption about the informational advantage of sellers over banks is from Mian and Smith (1992). Establishing cost advantages over competing lenders, these researchers refer to advantages in credit evaluation derived from regular contact with buyers and ongoing evaluations to produce information that can also be employed in credit extension decisions. This assumption was formalized in a theoretical model by Biais and Gollier (1997). Assuming that suppliers have private information about their customers, they argue that “trade credit can facilitate aggregation of the supplier's information with bank's information and thus alleviate an information asymmetry” (p. 904). More precisely, their three-agent model includes different types of buyers: good and bad. The buyer knows its own type, but the bank and supplier do not and have different signals about the type. Bank credit rationing arises because of imprecise signals on the buyer type when bank credit is the only source of financing and trade credit is not an option. By contrast, when sellers extend trade credit conditional on the receipt of a good signal from the buyer, the bank is induced to lend conditionally on their own positive signal. In this way, the information collected by the supplier is conveyed to the bank. Intuitively, if the seller is willing to bear the buyer's risk—extending credit in terms of delayed payments—the seller must have good information about the buyer. Thus, the authors conclude that “trade credit enables the private information of the seller to be used in the lending relationship and this additional information of the seller can alleviate credit rationing due to adverse selection” (p. 905).

Put differently, in the logic of the well-known delegated monitoring model of Diamond (1984), in which individuals invest in financial intermediaries to achieve economies of scale, the bank transfers the screening and monitoring process to firms that can more effectively carry out these tasks. Thus, trade credit is also a “result of the specialization and division of labour that occur within sophisticated financial markets” (Miwa & Ramseyer, 2008, p. 329). The informational comparative advantage of suppliers derives from their ability to collect information quickly and inexpensively because they are in the same industry as buyers (Jain, 2001) and have an existing day-to-day relationship. Suppliers are also better able to enforce non-collateralized debt repayment. The enforceability power of suppliers stems from the fact that they can threaten to stop supplying intermediate goods to their customers (Cunãt, 2007).

Empirically, Cook (1999) provides the first evidence of the signaling role of trade credit in periods characterized by structural changes in the Russian transition and a less developed economy. The findings allow the author to define trade credit as a screening device. Giannetti et al. (2011) find regularities between bank relationships and trade credit usage. They provide evidence of the signaling power of trade credit, which reveals favorable information to other lenders. However, their findings lead them to conclude that suppliers “enjoy an informational monopoly” in relation to “less informed banks” (Ibidem, p. 1288). Agostino and Trivieri (2014) test the informational advantage hypothesis of suppliers on the level of firm borrowing and provide evidence of the information content of trade credit for banks, especially when the relationship is weak, as measured by the increasing number of lenders and the fact that the bank cannot rely on soft information.

3 RESEARCH HYPOTHESES

Information asymmetry represents one of the most important frictions that can affect firms’ access to finance in all types of financial systems. On the one hand, in market-based financial systems, firms characterized by a high degree of information asymmetry prefer to obtain financing from banks rather than use public debt (Anderson & Makhija, 1999; Denis & Mihov, 2003). On the other hand, in bank-based financial systems, although firms use banks as the main source of financing, firms with a higher degree of information asymmetry may be credit-rationed by financial intermediaries (De Andrés Alonso et al., 2005; García-Teruel et al., 2014; Stiglitz & Weiss 1981). In these cases, constrained firms may have an incentive to use alternative sources of financing as trade credit because of suppliers’ better ability to overcome the screening, monitoring and enforcement problems faced by financial institutions (Love, 2011).

4 DATA AND SAMPLE FIRMS

Our data are from a proprietary loan-level database provided by one of the top five Italian banking groups. The database includes unique firm–bank identifiers, basic firm-level features and a set of loan-level characteristics, such as the results of loan requests (denied vs. approved).1 The dataset spans from 2009 to 2012 and is organized as an unbalanced panel. Moreover, all of the firms included in the sample are classified as SMEs, consistent with the European Commission's definition.2

These characteristics make the sample well-suited for analyzing relationship lending because Italian firms typically apply for loans at local bank branches instead of accessing bond or equity markets. Furthermore, none of the sample firms has issued public debt, and none are publicly listed. The size and legal characteristics of the sample firms and domestic banking laws ensure that banks do not have equity holdings in their client firms. For each borrower, we have information on the number and types of bank contacts per year, the credit rating, the contract terms, the share of firm debt provided by the bank, the total number of lenders and the date of the origination of the relationship with the bank.

Additionally, we use the Aida database managed by Bureau van Dijk to add some missing values to the balance sheet data. By matching the two datasets, we obtain an unbalanced panel of 47,105 firms that operated throughout Italy. Before obtaining the final dataset used in the regressions, we apply the following filters. Firms with missing values for any of the main variables are dropped. Moreover, the data related to continuous variables are winsorized at the 1% and 99% levels to address the problem of extreme outliers. In Table 1, we describe the variables used in the analysis and the relative summary statistics.

5 MULTIVARIATE ANALYSIS

Specifically, we consider trade credit (TC_it) as the ratio between trade payables and total debt, as well as other ratios, to check the robustness of the results. Because we assume that loan officers observe the past final year's volume, we also use the 1-year lagged value. Additionally, we scale payables by the firm's total assets (Tc*), as in Petersen and Rajan (1997) and other studies (e.g., Agostino & Trivieri, 2014; Garcia Teurel & Martínez-Solano, 2010), to address possible endogeneity concerns.

The vector B_it contains the bank assessment of firm creditworthiness. The credit analysis performed by the bank, typically a loan officer, aims to appraise the ability and willingness to repay the loan.

Credit analysts first use the economic prospectus and the borrower's past records, if available. The criteria used in the evaluation, known as the “five Cs of credit,” are capacity, capital, conditions, collateral and character (Thakor, 1995). Financial statements and the growth prospectus contain information useful for determining the borrower's capital and repayment capacity. Additional economic conditions, such as the macroeconomic environment or sector performance, also affect loan repayment. Furthermore, credit analysts are interested in the assets to be pledged as collateral that could easily be liquidated in the event of firm default. We use the internal rating as the overall evaluation of the firm's creditworthiness on a scale from 1 to 5; in the model, we include four rating dummies that take the value of 1 for the corresponding rating received and 0 otherwise. To avoid dummy traps, we do not include a rating 3 dummy.

The vector R_it considers the variables representing the relationship between the buyer and the bank, which enables the bank to gather information. One relationship variable is the length of the relationship between the firm and the bank (Duration). The length provides information on the strength of the borrower–lender relationship, and the unit of time is months. Another measure related to the strength of the relationship is its exclusivity between the firm and the bank, represented by the number of lenders that serve a firm (Lenders). Because a larger number of sources of financing may increase the bank's concerns over its ability to collect in the event of default (see Cole et al., 2004, p. 235), we expect a negative relationship (see also Cenni et al., 2015; Cole, 1998). Furthermore, we consider the number of banks as an inverse measure of the bank's information production (Carletti, 2004). Finally, the F component contains firm characteristics that could affect the firm's access to bank loans. In particular, the included control variables capture firm characteristics, such as firm Size; revenue Growth; South, if the firm is located in the southern part of Italy; Foreign, if the firm requests bank support for foreign operations (import and/or export), and sector, which can affect lending decisions by affecting the firm's quality, risk and creditworthiness (Aktas et al., 2012; Giannetti et al., 2011; Minetti et al., 2019; Ng et al., 1999; Petersen & Rajan, 1994, 1997; Summers & Wilson, 2002).

The magnitudes of the coefficients depend not only on the value of the covariates, x_t (equation 1) but also on the value of the unobserved heterogeneity, c_i. To estimate the partial effects at average and average partial effects (Wooldridge, 2010), we assume strict exogeneity conditional on the unobserved effects, c_i (Chamberlain, 1984), and run a panel probit regression with random effects in Section 6.1. We then relax this assumption considering the possible endogeneity in Section 6.2 and allow for any serial correlation with the pooled probit estimator with clustered standard errors in Section 6.4.

6 RESULTS AND DISCUSSION

6.1 Signaling role of trade credit in bank lending decisions

In this section, we focus on the central research question of this study: Does trade credit play a role in lending decisions? Before we discuss the multivariate analysis based on equation (1), we preliminarily investigate the relationship between trade credit and firm creditworthiness because creditworthiness represents one of the most important dimensions considered in banks’ lending decisions. To that end, we perform a mean equality test to compare the level of trade credit and other variables related to different classes of risk, and we report the results in Table 2.

TABLE 2. Two-group mean comparison t-test

Variable	Class of risk = 1 (low) group a		Class of risk ≠ 1 group b		t-test mean (b)–mean (a)
	Obs.	Mean	Obs.	Mean
Accpay	24,952	452.24	38,172	404.58	−47.65***
Debt	24,952	1251.23	38,172	1322.31	71.08***
Tc	24,952	0.37	38,172	0.31	−0.05***
Duration	24,952	4.53	38,172	3.83	−0.70***
Age	24,952	16.77	38,172	12.35	−4.41*
Growth	24,952	0.08	38,172	0.22	0.13***

	Class of risk = 2 group a		Class of risk > 2 group b		t-test mean (a)–mean (b)
Accpay	18,710	404.73	19,462	404.44	−0.28
Debt	18,710	1257.98	19,462	1384.16	126.17***
Tc	18,710	0.33	19,462	0.30	−0.02***
Duration	18,710	3.99	19,462	3.69	−0.29**
Age	18,710	13.40	19,462	11.34	−2.06
Growth	18,710	0.17	19,462	0.26	0.08***

	Class of risk = 3 group a		Class of risk > 3 group b		t-test mean (a)–mean (b)
Accpay	11,941	392.64	7521	423.19	30.55***
Debt	11,941	1344.95	7521	1446.41	101.45***
Tc	11,941	0.30	7521	0.29	−0.01***
Duration	11,941	3.65	7521	3.74	0.09**
Age	11,941	11.78	7521	10.64	−1.14
Growth	11,941	0.27	7521	0.24	−0.03**

	Class of risk = 4 group a		Class of risk > 4 group b		t-test mean (a)–mean (b)
Accpay	6646	425.97	875	402.13	−23.83
Debt	6646	1452.85	875	1397.48	−55.36
Tc	6646	0.29	875	0.28	−0.014**
Duration	6646	3.72	875	3.95	0.23**
Age	6646	10.76	875	9.69	−1.07
Growth	6646	0.26	875	0.08	−0.18**

	Class of risk = 5 (high) group a		Class of risk < 5 group b		t-test mean (a)–mean (b)
Accpay	875	402.13	62,249	423.72	21.59
Debt	875	1397.48	62,249	1292.76	−104.72**
Tc	875	0.28	62,249	0.34	0.057***
Duration	875	3.95	62,249	4.11	0.16**
Age	875	9.69	62,249	14.16	4.47
Growth	875	0.08	62,249	0.17	0.08**

• Note: This table reports a two-group mean comparison t-test performed to compare the average difference between each combined group (by risk) with the group of remaining observations with higher risk. For example, for risk class 1 (risk class 2), the average values of accounts payable, TC and total debt are compared (in terms of difference) with the average values reported by firms in classes 2–5 (3–5). The average values of firms in class 5 are compared with all remaining observations. Class 1 is the lowest risk class, whereas class 5 is the highest.
• ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.

The ratio between account payables and total debt decreases with risk, such that the higher the risk of the borrower, the lower the trade credit ratio. This finding is not surprising because trade credit may be considered a reflection of firms’ reputation. Based on this analytical perspective, a borrower's increasing level of debt may induce suppliers to reduce the credit that they extend to the borrower (as shown in Figure 1 in the comparison of Accpay, Debt and Tc under different risk levels).

Furthermore, the comparison between age and risk shows that the former is inversely related to the latter because riskier firms are younger (see Figure 1). This evidence may reflect the effect of banks’ attitudes regarding holding established relationships with safer firms, or it may be a consequence of the higher level of (positive) hard and soft information available for mature firms, which positively affects the borrower's rating. Then, we examine the correlation between trade credit and age based on a graphical analysis and report the results in Figure 1. Interestingly, although account payables increase with age, their value scaled by debt (Tc) is higher among younger firms and lower among older firms because of the high total debt of the latter.

After this preliminary investigation, we examine the results of our main regression based on equation (1) and analyze the contribution of trade credit to the loan approval process. We employ Tc (the ratio between account payables and total debt) as the dependent variable for several reasons. The ratio of account payables to total debt captures the importance of trade credit in the financing of a firm's assets. Thus, the loan officers at the bank from which we collected the data are typically prone to consider the value of trade credit in relative terms (with respect to total debt) because the reporting scheme of the balance sheet that they employ is mainly oriented to note the relative values of each balance sheet item.

The results of the multivariate analysis are reported in Table 3 and generally provide evidence of the signaling role of payables in loan approvals. Consistent with our hypothesis, the bank choice of approval or denial seems contingent on the extension of credit from suppliers to buyers, ceteris paribus.

TABLE 3. Panel probit estimation of loan approval decisions

	(I)	(II)	(III)	(IV)	(V)	(VI)	(VII)	(VIII)
Variable	Coefficients	Marginal effects	Coefficients	Marginal effects	Coefficients	Marginal effects	Coefficients	Marginal effects
Tc	0.2982***	0.0227***	–	–	–	–	–	–
	(0.055)	(0.004)
Tct–1	–	–	0.2926***	0.0229***	–	–	–	–
			(0.051)	(0.004)
Tc*	–	–	–	–	0.1346**	0.0102**	–	–
					(0.062)	(0.005)
Tc*t–1	–	–	–	–	–	–	0.1723***	0.0135***
							(0.059)	(0.005)
Extreme	−0.1124***	−0.0085***	−0.0921***	−0.0072***	−0.0856***	−0.0065***	−0.0791***	−0.0062***
	(0.022)	(0.002)	(0.021)	(0.002)	(0.022)	(0.002)	(0.021)	(0.002)
R1	0.2965***	0.0226***	0.3045***	0.0239***	0.3149***	0.0239***	0.3211***	0.0252***
	(0.030)	(0.002)	(0.030)	(0.002)	(0.030)	(0.002)	(0.030)	(0.002)
R2	0.1151***	0.0088***	0.1263***	0.0099***	0.1220***	0.0093***	0.1332***	0.0104***
	(0.028)	(0.002)	(0.028)	(0.002)	(0.028)	(0.002)	(0.028)	(0.002)
R4	−0.3791***	−0.0288***	−0.3539***	−0.0277***	−0.3825***	−0.0290***	−0.3561***	−0.0279***
	(0.030)	(0.002)	(0.031)	(0.003)	(0.030)	(0.002)	(0.031)	(0.003)
R5	−0.8043***	−0.0612***	−0.7840***	−0.0614***	−0.8121***	−0.0617***	−0.7920***	−0.0621***
	(0.055)	(0.004)	(0.056)	(0.005)	(0.055)	(0.004)	(0.056)	(0.005)
Duration	0.0002	0.0000	0.0002	0.0000	0.0001	0.0000	0.0001	0.0000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Lenders	−0.0247***	−0.0019***	−0.0237***	−0.0019***	−0.0287***	−0.0022***	−0.0268***	−0.0021***
	(0.004)	(0.000)	(0.005)	(0.000)	(0.004)	(0.000)	(0.004)	(0.000)
Size	−0.0397***	−0.0030***	−0.0389***	−0.0030***	−0.0257**	−0.0019**	−0.0291**	−0.0023**
	(0.011)	(0.001)	(0.011)	(0.001)	(0.011)	(0.001)	(0.011)	(0.001)
Growth	0.0141	0.0011	0.0216*	0.0017*	0.0162	0.0012	0.0224*	0.0018*
	(0.012)	(0.001)	(0.012)	(0.001)	(0.012)	(0.001)	(0.012)	(0.001)
Foreign	0.0485	0.0037	0.0732*	0.0057*	0.0487	0.0037	0.0714	0.0056
	(0.042)	(0.003)	(0.044)	(0.003)	(0.042)	(0.003)	(0.043)	(0.003)
South	−0.0986***	−0.0075***	−0.1009***	−0.0079***	−0.0973***	−0.0074***	−0.0968***	−0.0076***
	(0.021)	(0.002)	(0.022)	(0.002)	(0.021)	(0.002)	(0.022)	(0.002)
Constant	2.0840***		2.0358***		2.0520***		2.0276***
	(0.073)		(0.073)		(0.072)		(0.073)
Observations	63,132	63,132	59,250	59,250	63,136	63,136	59,284	59,284

• Note: This table shows the panel probit estimation for bank lending decisions and the predicted probabilities based on the model below:
  $$\protect\begin{align*} Pr{\rm{ }}(Approva{l_{it\ }} = \ 1|{\rm{ }}{x_{it}})\ = \ \Phi ({\beta _0} + {\rm{ }}{\beta _1}T{C_{it}} + {\beta _2}{B_{it}} + {\rm{ }}{\beta _3}{R_{it}} + {\rm{ }}{\beta _4}{F_{it}} + {c_i}). \protect\end{align*}$$
  The vector B_it contains the bank assessment of firm creditworthiness, the vector R_it contains relationship lending variables and F_it contains firm control variables_. The coefficients are reported in odd-numbered columns, and the average marginal effects are reported in even-numbered columns. The dependent variable is Approval, which is a dichotomous variable for bank lending decisions that takes the value of 1 if the bank approves the request for financing of firm i in year t and 0 otherwise. Tc_it is the ratio between trade payables and total debt; we also use the 1-year lagged value, TC_it–1. Extreme is a binary variable that takes the value of 1 if firm i has an average payables term higher than the average duration of the sector to which it belongs. We include four rating dummies that take the value of 1 for the corresponding rating received and 0 otherwise. To avoid dummy traps, we do not include the rating 3 dummy. Duration is the length of the relationship between the firm and the bank in months. Lenders is number of lenders that serve a firm. Size is the log value of the total revenue of firm i at the end of year t. Growth is the percentage growth of the total revenue of firm i between the end of year t–1 and the end of year t. South is a binary variable that takes the value of 1 if firms are located in southern regions and 0 otherwise. Foreign is a binary variable that takes the value of 1 if firm i has purchased from the bank in year t at least one banking service related to import/export trading. Additionally, we scale payables by the firm's total assets (Tc*) and its lagged value (Tc*_it–1) in columns V–VIII. All independent variables are defined in detail in Table 1, and the t-statistics are in parentheses.
• ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.

Approval is used as the dependent variable in the probit model discussed in Section 4, and the first column in Table 3 tests whether the presence of trade credit contains valuable credit quality information for banks and contributes to lending decisions. The coefficient of trade credit is positive and significant at all conventional levels. Because we assume that banks observe suppliers’ behavior ex-post (i.e., loan officers review the balance sheet from the most recent available year), we also consider the lagged value of Tc and note that the results remain unchanged. Thus, banks seem to incorporate the use of trade credit when granting loan decisions. The positive role of trade credit between buyers and suppliers in lending decisions is consistent with the models of Emery (1984) and Biais and Gollier (1997). The informational advantage of a supplier that collects information during day-by-day business relationships is recognized by the bank and factors into its decision. This evidence is useful for understanding how agreed-upon delayed payments affect bank decisions. Trade partners can serve not only as a signaling device but also as a means of monitoring because lenders, as trade partners, know the industry well and reduce their risk by closely monitoring borrowers (Miwa & Ramseyer, 2008).

The results seem to support H₁: Bank lending decisions are influenced by trade credit, and trade credit plays a signaling role in the loan approval process. In particular, trade credit increases the likelihood of a firm to obtain a positive response to a loan request, ceteris paribus. Thus, banks seem to use trade credit information in lending decisions to efficiently allocate capital. The informational value derived from commercial credit is not assumed sic et simpliciter, but the bank uses and processes this information proactively, increasing the granularity of the information.

To gain deeper insights into the monitoring role of suppliers, we include in our regression the duration of payables in days. Specifically, based on previous studies (e.g., Giannetti et al., 2011), we include the binary variable Extreme, which indicates whether a firm has an average payables term with suppliers that is longer than the average duration of the sector to which it belongs (Extreme = 1) or not (Extreme = 0). The results of this inclusion show that a duration longer than the median negatively impacts banks’ final loan approval decisions (columns I–VIII, Table 3). One possible reason for this finding is that a greater extension of credit terms (Extreme = 1) could be symbolic of the liquidity tension of a firm, which could negatively affect loan granting decisions.

Not surprisingly, the bank's valuation of the borrower's creditworthiness, as determined by its internal rating system, affects loan approval decisions. Firms that belong to the least risky classes (R1 or R2) have a higher probability of obtaining a loan. Conversely, if the firm requesting financing receives a riskier valuation class (R4 or R5), the request is less likely to be granted.

The analysis of the contribution of trade credit to lending decisions also considers the effect of the repeated interactions between the bank and the firm because past experience with the firm could affect the denial or approval of a loan request. Before disentangling the role in the lending decision played by the presence of trade credit with pooled information gathered by the bank during its relationship with the firm, we briefly analyze the regression results of the relationship lending coefficients in Table 3. From the firm's perspective, a strong bank relationship is considered a valuable asset because it reduces the cost and increases the availability of credit.

In our main specification, we include the number of relationships that a firm has with financial intermediaries (Lenders) and the duration of the bank–firm relationship (Duration) to assess the strength and intensity of relationships as a source of soft information production and the effect on the probability of obtaining a loan. The negative coefficient (statistically significant at the 1% level) of Lenders in all specifications in Table 3 suggests that increasing the number of firm financial intermediary relationships reduces the probability of obtaining loans. Consistent with previous studies (e.g., Carletti, 2004), multiple lending relationships may represent an inverse measure of soft information production. Thus, the higher the number of lenders is, the lower the accumulation of soft information obtained by banks and, consequently, the lower the probability of a borrower establishing reliability using qualitative information.

Moreover, the relationship between Approval and Duration shows that firms involved in long-lasting relationships with banks are more likely to be approved, although this result was not statistically significant. In addition, firm characteristics, such as size (Size) and location (South), negatively affect the probability of obtaining a loan.

Because participation in a global supply chain can allow firms to receive liquidity from each other by allowing for the delayed payment of products (Minetti et al., 2019), we control our results by including Foreign, which captures the buyer's participation in the supply chain (without any information on the intensity of this participation). However, this variable does not seem to play a significant role in explaining the probability of the loan approval decision, and the other control variables seem to be statistically irrelevant (i.e., firm growth).

We estimate the average marginal effect to compute the impact and weight of each regressor (columns II, IV, VI and VIII of Table 3). As noted by Cameron and Trivedi (2005), in the non-linear model, the average behavior of individuals differs from the behavior of the average individual (marginal effects at the mean). This type of estimation should lead to results similar to those obtained based on ordinary least squares. From the previous regression (odd columns), we retain all of the effects of the variables in the model because the signs remain. This analysis allows us to discuss the magnitude of the coefficients with respect to the probability.

An increase of 1% in the ratio of payables to total debt increases the probability by 2.27% (see column II, Table 3), and a similar result is observed for previous year-end observations (Tc_t–1, columns VI and VIII, Table 3). Although firms benefit from the use of trade debt, extra delayed payments lower the probability of financing by 0.8% (column II, Table 3). Less risky firms (belonging to risk classes 1 and 2, i.e., R1 and R2) have the highest probability of requests being granted, whereas risky firms (belonging to risk classes 4 and 5, i.e., R4 and R5) show the opposite effect; specifically, the classification of a firm in the worst risk class (R5) reduces the probability of loan approval by approximately 6%, on average (columns II, IV, VI and VIII). These results reflect the tendency of banks to grant loans to low-risk borrowers to improve their allocative efficiency and preserve capital as a result of institutional changes, such as Basel II and the aftermath of the financial crisis.

7 TRADE CREDIT AND CORPORATE FAILURE

The results shown thus far tell a story in which trade credit plays a beneficial role in the loan approval process, even if the effect is reduced in the presence of a close relationship with the bank (probably because of a substitution effect of soft information). Accounts payable can act as transmitters of a good reputation signal, while trade credit is another component of a company's debt and, as such, can be related to the company's bankruptcy. When a firm approaches bankruptcy, it is possible to identify changes in the supply of trade credit rather than changes in the demand for credit, depending on switching costs (Garcia-Appendini & Montoriol-Garriga, 2020). A trade debtor's failure may generate significant trade credit losses for suppliers and may be a trigger for a propagation mechanism such that debtor failures lead to creditor failures (Jacobson & Von Schedvin, 2015).

However, trade credit also represents a determinant of corporate failure given the buyer–seller collusion suggested by Biais and Gollier (1997). Specifically, the case of collusion between the seller and the buyer is represented by the potential attempt of the seller “to extend credit to bad buyer[s] even if this exposed him to a large credit risk because, contingent on the supplier extending trade credit, the bank would provide complementary financing” (Ibidem, p. 913). Hence, on the basis of the distorted signal, the bank erroneously attributes a good rating to the buyer. At this point, the buyer may reveal/disclose its type during the relationship, resulting in a non-performing loan. In contrast, the seller shares the surplus extracted by the bank with the buyer in the form of inflated prices or services.

In an attempt to disentangle this possible effect and, in a stricter sense, the role of trade credit in corporate failure, we consider an additional specification to observe the possible effect of trade credit on non-performing loans. In particular, as a measure of corporate distress, we employ the Default_it variable, which takes the value of 1 if the ith firm has a non-performing loan in year t. Then, we apply a probit regression model to the panel of firms receiving a new loan in year t–1 and measure the marginal effect of trade credit in years t–1 and t–2 on firm failure risk.

The regression results are reported in Table 8. For both periods (t–1 and t–2), the average marginal effect for the TC variable is positive and significant, indicating that trade credit is associated with, on average, an increased likelihood of trade debtor failure in subsequent years (see columns II and IV of Table 8). Taken together with the results of the previous sections (where trade credit appears to facilitate loan approvals) and considering the low number of distressed firms in the sample, these findings do not allow us to exclude the presence of false signals in trade credit and the competitive edge that trade creditors have over banks in their ability to assess prospective debtors’ creditworthiness and ongoing activities (see, e.g., Biais & Gollier, 1997; Burkart & Ellingsen, 2004).

8 CONCLUSION

Trade credit is a signaling device for banks because suppliers bearing the risk of a credit extension lead banks to grant additional financing to firms. Theoretical papers (e.g., Biais & Gollier 1997; Burkart & Ellingsen, 2004) find that trade credit ameliorates the asymmetric information between banks and firms that can lead to credit rationing. Biais and Gollier (1997) suggest that trade credit may also act as a proxy for firm creditworthiness. We extend this conjecture and suggest that trade credit may play a role in the loan approval process because it represents a signal of the buyer's reputation in the market. Consistently, it may be beneficial for firms requesting loans, especially if the firms are opaque. Our main result confirms this state of affairs and suggests a new form of complementarity between bank loans and trade credit.

We interpret this result along the lines of the signaling approach of trade credit, which suggests that the availability of trade credit helps in obtaining access to bank debt and represents trade credit as a tool used to obtain bank credit. However, the role of trade credit is not risk-free because false signals cannot be excluded. In particular, because trade credit is a form of debt, its prolonged and increasing recourse can negatively affect the probability that the company will default. Thus, trade credit can have more than one meaning, and banks should not be limited to its disclosure in balance sheet data and should obtain more details regarding the type of inter-firm network, cost and, finally, duration of the relationship with the supplier.