Trade Credit and the Transmission of Unconventional Monetary Policy

Abstract

Conventional monetary policy affects access to external finance through its effects on bank lending and consequently affects primarily small and bank-dependent firms (Gertler and Hubbard 1988; Gertler and Gilchrist 1994; Holmstrom and Tirole 1997). Large firms with access to public debt markets are less sensitive to the monetary policy stance.

Unconventional monetary policy involves directly purchasing assets in public debt markets in order to lower interest rates and stimulate the economy. Central banks purchase short-term and long-term Treasury bonds and mortgage-backed securities, which can benefit firms indirectly (Foley-Fisher, Ramcharan, and Yu 2016), but more recently also target nonfinancial firms directly through corporate bond purchases. One such program is the European Central Bank’s (ECB) Corporate Sector Purchase Program (CSPP), which involves the purchase of investment-grade corporate bonds. These large-scale asset purchases affect bond yields and issuance volumes (Abidi and Miquel-Flores 2018; Galema and Lugo 2019; Zaghini 2019; Todorov 2020) and benefit primarily large firms with access to bond markets. Small firms can benefit indirectly from these purchases as large firms’ demand for bank loans decreases and banks expand credit supply to small firms (Grosse-Rueschkamp, Steffen, and Streitz 2019; Arce, Mayordomo, and Gimeno 2020).

This paper proposes and tests a new channel through which the benefits of unconventional monetary policy can be redistributed to firms without access to bond markets: the trade credit channel, which operates independently from any effects of monetary policy on bank lending. Using the ECB’s CSPP as a laboratory, we hypothesize that firms with bonds eligible under the CSPP act as financial intermediaries by providing trade credit to their customers as eligible firms experience a decrease in the cost of capital. The ability to issue bonds at a relatively lower yield leads these firms to ration fewer marginal customers from trade credit. We examine whether this channel mitigates or exacerbates asymmetries in the transmission of monetary policy by studying which customers are supported by suppliers with CSPP-eligible bonds and the resultant real effects. We also explore how the CSPP affects eligible firms’ power to acquire and retain customers and thus their competitive position in product markets.

The CSPP was announced in March 2016 and implemented starting in June 2016 and it was designed with a strict eligibility criteria in place, as only investment-grade bonds issued by nonfinancial firms from the euro area can be purchased by the ECB. The design of the CSPP allows us to implement a difference-in-differences analysis to address concerns that concurrent shocks to the implementation of the CSPP can affect eligible firms and their customers.

Cash flows from the CSPP are disproportionately injected into the core of the euro area, including France, Germany, and the Benelux Economic Union (Bayoumi and Eichengreen 1993), where most firms with eligible bonds are located. In fact, firms in core countries are able to issue more bonds thanks to more developed capital markets and stronger legal institutions (Becker and Josephson 2016). However, if targeted (large and financially unconstrained) firms pass on the additional liquidity to their customers through trade credit, unconventional monetary policy tools can also reach small and financially constrained firms through production networks. In addition, while firms located in core countries may experience a larger direct effect from the CSPP through bond issuance and lower cost of capital, whether the increase in liquidity spills over to other regions depends on the geography of production networks.

We investigate these questions using new data containing information on firm-level customer-supplier networks. We compare the amount of trade credit extended by eligible firms and noneligible firms, before and after the CSPP announcement. We find that eligible firms increase the amount of trade credit they provide to customers (i.e., accounts receivable) more than noneligible firms following the CSPP. As we would expect, at the same time, the accounts payable of customers of eligible firms increase after the CSPP.

We find no statistically significant differences in trade credit between eligible firms, or their customers, and control firms before the start of the CSPP. In addition, we find a discrete increase in eligible firms’ accounts receivable around the investment-grade rating cutoff, the most relevant criterion for eligibility under the CSPP. This suggests that changes in trade credit are not triggered by demand shocks or other contemporaneous shocks such as, for example, worsening credit conditions that could have simultaneously increased firms’ reliance on trade credit and led the ECB to start the CSPP. We provide evidence that the effects of the CSPP are instead driven by the decrease in the cost of external funding triggered by the ECB’s asset purchases. Indeed, we find that eligible firms that provide more trade credit issue bonds after the start of the CSPP.

One important feature of our setting is that eligible firms are, by the nature of the program, larger than noneligible firms, on average. To account for the possibility of preexisting differences between treated and control firms, we repeat our tests using matched samples. We also sort firms into size deciles and include size decile dummies interacted with year dummies in the regressions. We find similar estimates in these more stringent specifications, which indicates that asymmetric shocks affecting firms of different sizes are unlikely to drive our findings.

The effects of the trade credit channel of monetary policy are heterogeneous across regions and firms. Firms targeted by the CSPP extend trade credit to customers that are more financially constrained. In particular, we find that smaller firms, non-investment-grade firms, unrated firms, firms with higher leverage and firms with lower tangible assets to pledge as collateral receive more trade credit.

The CSPP also produces redistributive effects across regions. While the ECB aimed to purchase investment-grade bonds of firms in any euro area country, firms in core countries were able to issue more bonds after the announcement of the program than firms in periphery countries. Arguably as a consequence of the asymmetric change in financial conditions, we only detect an increase in accounts receivable for suppliers located in the core and no significant effects in the periphery. We find the opposite when we consider the customers of eligible firms. Customers located in core countries show small and insignificant increases in accounts payable, while customers in periphery countries show a significant increase in accounts payable. We provide direct evidence that links between suppliers in the core and customers in the periphery drive the effects. These results suggest that trade credit helped to relax financial constraints in periphery countries where banks were more affected by the 2010–2011 European sovereign debt crisis. Monetary policy transmission through production networks mitigates the asymmetric effects that arise from the regional distribution of eligible firms and their ability to issue investment-grade bonds, which benefits core countries to a larger extent.

Finally, we show that the trade credit channel produces real effects. As a result of the increase in accounts payable, the customers of eligible suppliers increase long-term investment in fixed assets and human capital. They also increase investment in working capital by extending more trade finance to their own customers and increasing inventories. At the same time, eligible firms improve their position in product markets. Eligible suppliers acquire new customers and are more likely to maintain existing business relationships, thus increasing their market share as a result of the CSPP.

Our findings highlight mechanisms of tantamount importance in light of the expanded direct assets purchases announced by the Federal Reserve Board and the ECB in March 2020. Notably, both the unlimited quantitative easing program of the Federal Reserve Board and the Pandemic Emergency Purchase Program (PEPP) of the ECB involve direct interventions in corporate bond markets. We highlight a new channel through which quantitative easing affects the real economy and complements the stimulus arising from the effects of asset purchases on banks’ balance sheets and lending. Trade credit can transmit the stimulus of unconventional monetary policy interventions to firms that are not directly targeted by the policy. However, monetary policy interventions may promote concentration in upstream industries with long-run consequences on industrial structure.

Our paper contributes to the literature on the effects of large-scale asset purchases on bank lending and real economic activity (Rodnyansky and Darmouni 2017; Acharya et al. 2019; Grosse-Rueschkamp, Steffen, and Streitz 2019; Chakraborty, Goldstein, and MacKinlay 2020; Di Maggio, Kermani, and Palmer 2020).

We also contribute to the literature on trade credit. Previous research provides mixed evidence on whether trade credit attenuates or amplifies the transmission of conventional monetary policy to the real economy. While Gertler and Gilchrist (1993) find no substitution between bank loans and trade credit following a tightening of monetary policy, Nilsen (2002) provides evidence that both small and large firms increase trade credit following monetary contractions. Other studies explore whether trade credit can provide an alternative source of liquidity that mitigates the effects of bank liquidity shocks. Love, Preve, and Sarria-Allende (2007) find that trade credit collapsed in the aftermath of the 1997 Asian crisis, while Garcia-Appendini and Montoriol-Garriga (2013) find that cash-rich suppliers extended more trade credit during the 2007–2009 global financial crisis. Restrepo, Cardona Sosa, and Strahan (2019) show that firms rely less on short-term loans and more on cash and trade credit for liquidity management following an exogenous increase in the relative cost of short-term bank credit in Colombia in 2011.

Last, we contribute to the growing literature on the importance of production networks in the transmission of economic shocks. A number of papers explore how negative shocks are transmitted through the supply chain and show that upstream negative liquidity shocks are amplified and transmitted to customers (Boissay and Gropp 2013; Jacobson and von Schedvin 2015; Barrot and Sauvagnat 2016). A few recent papers consider the role of bank liquidity shocks in supply chains (Costello 2020; Huremovic et al. 2020; Alfaro, García-Santana, and Moral-Benito 2021) and how banking structure is related to the propagation of shocks along the supply chain (Giannetti and Saidi 2019). Using the stock market reaction to monetary policy shocks, Ozdagli and Weber (2019) show that input-output linkages are an important transmission mechanism of macroeconomic shocks through higher-order demand effects.

To the best of our knowledge, we are the first to study the trade credit channel of unconventional monetary policy, which is independent of the bank lending channel. By exploiting exogenous variation in suppliers’ ability to access external debt financing, we show that trade credit enhances the transmission of quantitative easing interventions to the real economy. We also contribute to a nascent strand of the literature that explores regional heterogeneity in the transmission of monetary policy (Beraja et al. 2018) by studying how the effects of large-scale asset purchases are redistributed across regions through production networks.

1. Data and Methodology

This section describes the data, variables, and the empirical methodology.

1.1 Sample

Our initial sample consists of a panel of publicly listed and privately held firms in the period 2013–2017 drawn from the Bureau Van Dijk’s Orbis database, which contains financial statements for companies worldwide.¹ We restrict the sample to firms based in the 19 member states that are part of the Economic and Monetary Union of the European Union (euro area). We exclude from our sample firms that are classified as small companies by Orbis.² We also exclude financial firms (SIC codes 6000–6999) and public administration entities (SIC codes 9000–9999). Finally, we require nonmissing data on the main firm outcomes: the ratio of accounts receivable to sales (Accounts receivable), and the ratio of accounts payable to sales (Accounts payable).³ The control variables include total assets (Assets), the ratio of cash to total assets (Cash), the ratio of property, plant, and equipment to total assets (PPE), the ratio of net income to sales (Net margin), and the ratio of total liabilities to total assets (Liabilities).

We also consider additional firm outcomes for investment in short-term assets, investment in assets and capital expenditures, employment, and financing decisions. The investment variables include the change in accounts receivable over lagged total assets (⁠|$\Delta $|Accounts Receivable), the change in inventories over lagged total assets (⁠|$\Delta $|Inventories), the change in total assets over lagged total assets (Asset growth), the change in tangible assets plus depreciation over lagged total assets (CAPEX), and the change in the number of employees over lagged number of employees (Labor growth). The financing variables include the change in accounts payable (⁠|$\Delta $|Accounts payable), the change in financial debt (⁠|$\Delta $|Total debt), the change in long-term debt (⁠|$\Delta $|Long-term debt), the change in short-term debt (⁠|$\Delta $|Short-term debt), and the change in cash and equivalents (⁠|$\Delta $|Cash); all variables are scaled by lagged total assets.

Table 1 reports summary statistics for the variables used in our analysis. The final sample consists of 510,298 unique firms for a total of 2,248,514 firm-year observations. Panel A of Table IA.1 in the Internet Appendix lists the number of observations and firms by country. Table A.1 in the appendix defines the variables.

1.2 CSPP and eligible firms

The ECB started a series of direct asset purchases programs to ease monetary conditions in the euro area, lower long-term interest rates and ultimately increase long-term inflation expectations and achieve the inflation target in 2012 (Rostagno et al. 2021). Initially, the outright monetary transmission program and the asset purchase program were limited to asset-backed securities and sovereign bonds. To further ease credit conditions in the euro area, on March 10, 2016, the ECB announced the CSPP, which implied an expansion of its asset purchase program to include investment-grade corporate bonds, as a tool to strengthen its accommodative monetary policy stance and to improve firms’ financing conditions.⁴

We start from the list of marketable bonds accepted as collateral for Eurosystem credit operations that was published by the ECB the day before the CSPP announcement on March 9, 2016. From this list, consistent with the CSPP eligibility criteria, we retain euro-denominated securities (denominations EUR, DEM, FRF) classified as bonds (type AT01) or medium term notes (type AT02) issued by corporations (issuer group IG3) and financial corporations other than credit institutions (issuer group IG9) resident in a country member of the euro area.⁵

To assign each bond to a unique firm, we first consider the issuer name of each bond in the list of bonds accepted as collateral by the ECB. We also consider the name of corporations and financial corporations other than credit institutions (guarantor groups GG3 and GG9) that guarantee eligible bonds provided that the guarantors are resident in a country member of the euro area. In fact, several bonds are issued by financial subsidiaries and are guaranteed by the parent organization (e.g., bonds issued by “Volkswagen Intl Finance N.V.” and by “Iberdrola Finanzas S.A.U.” are guaranteed by “Volkswagen AG” and by “Iberdrola S.A.,” respectively). Next, we apply a fuzzy-string matching technique to identify in Orbis the eligible firms (i.e., firms with bonds eligible for purchase under the CSPP before the announcement). As a result, our sample includes 151 unique nonfinancial eligible firms for which we have data on the variables of interest. Panel A of Table IA.1 in the Internet Appendix reports the number of eligible firms by country. Panel B lists the frequency of firms with different ratings, distinguishing between eligible and noneligible firms.

Figure 1 shows that the percentage that each country represents of the gross domestic product (GDP) in the euro area as of 2015 (panel A) and of the total number of eligible bonds (panel B) and amount outstanding (panel C) in the euro area just before the start of CSPP (as of March 2016). We find that about 77|$\%$| of the universe of eligible bonds under the CSPP are issued by firms in core countries (Austria, Belgium, Finland, France, Germany, Luxembourg, and Netherlands), while countries in the core represent only about 66|$\%$| of the GDP of the euro area.⁶

Figure 1

Eligible corporate bonds under the CSPP by country

Panel A reports the percentage that each country represented of the GDP in the euro area as of 2015. Panels B and C show the percentage that each country (i.e., country of risk) represented of the ECB’s CSPP in terms of the number of eligible bonds and total amount outstanding in the euro area as of March 2016. The sample consists of bond issues of nonfinancial firms based in the euro area.

Open in new tabDownload slide

Previous literature has shown that the CSPP fostered the ability of eligible firms to tap public debt markets stimulating new issuance of bonds and reducing the bond yields of eligible firms (Abidi and Miquel-Flores 2018; Galema and Lugo 2019; Grosse-Rueschkamp, Steffen, and Streitz 2019; Zaghini 2019; Arce, Mayordomo, and Gimeno 2020; Todorov 2020).

The geographic distribution of eligible firms implies that the consequences of the CSPP in terms of firms’ ability to issue investment-grade bonds and benefit from lower yields are asymmetric across regions. Panel A of Figure 2 shows the issuance of investment-grade euro-denominated bonds before and after the CSPP. In core countries, the new issuance of investment-grade euro-denominated bonds increased by €51 billion in 2016 (from €76 billion to €127 billion). New issuance of investment-grade euro-denominated bonds in periphery countries increased by €11 billion (from €18 billion to €29 billion). The figure also shows that the increase in issuance for investment-grade bonds continues through 2017.

Figure 2

New issuance of corporate bonds around the CSPP: Core versus periphery countries

This figure shows the amount of bonds issued (€ billion) by core and periphery euro area firms in the euro-denominated corporate bond market. Bond issuance data are obtained from SDC New Issues and includes bond issues of nonfinancial firms based in the euro area in the 2013–2017 period. Panel A shows new issuance of investment-grade bonds. Panel B shows new issuance of non-investment-grade bonds.

Open in new tabDownload slide

While differences in bond issuance also could be explained by differences in firm-specific credit demand, panel B of Figure 2 shows that the new issuance of non-investment-grade euro-denominated bonds (that were not targeted by the CSPP) was almost unchanged from 2015 to 2016 in both core and periphery countries. Given that before the CSPP investment-grade bonds and non-investment-grade bonds show similar behavior, differences in the demand for credit are unlikely to explain the differences in bond issuance.

1.3 Customers of eligible firms

We link each eligible firm (i.e., supplier) to all disclosed customers reported in the FactSet Revere Supply Chain Relationship database. FactSet Revere collects customer and supplier relationship information from primary public sources, such as SEC 10-K annual filings, investor presentations, and press releases. If we consider customer and supplier relationships, FactSet Revere include over 25,000 companies worldwide, which are the source of the supply chain relationship data, and over 105,000 target companies, which are disclosed by source companies.

Using FactSet Revere, we can track the effects on suppliers and their customers. We identify customers of eligible firms using direct and reverse relationships. A direct relationship is disclosed by the company that lists the target company as a material customer, and a reverse relationship is disclosed by another company listing the source company as a material supplier. As a result, our data include a comprehensive network of supply chain interconnections.

To better understand our data, consider “Deutsche Telekom AG” as an example. Using direct relationships, the company discloses 24 active material customers as of 2015. This list includes public entities, such as “Government of Germany,” “Government of Switzerland,” and “European Commission,” and corporations, such as “Deutsche Post AG,” “Daimler AG,” “ABB Ltd.,” and “Netflix, Inc.” Using reverse relationships, “Deutsche Telekom AG” is disclosed as a material supplier by an additional 11 unique companies. Customers that reported the firm as a supplier include “Freenet AG,” “Drillisch AG,” and “KION Group AG,” among other corporations. In total, considering both direct and reverse relationships, “Deutsche Telekom AG” has a total of 35 unique customers identified using FactSet Revere.

We begin by filtering the FactSet Revere data only selecting suppliers that are CSPP-eligible firms. On average, eligible firms report 7.6 customers domiciled in the euro area. We then match the list of customers of eligible firms in FactSet Revere to Orbis using ISIN identifiers, when available, and a fuzzy-string matching algorithm using names for the remaining firms. The final sample includes 318 unique nonfinancial customers of eligible firms, domiciled in the euro area, after matching with Orbis to obtain firm fundamentals.

Panel A of Table IA.1 in the Internet Appendix reports the number of customers of eligible firms by country. On average, customers that we identify through direct and reverse relationships (i.e., from the reporting of customers or suppliers) have similar size. While the customers of eligible firms in our sample are, on average, smaller than their suppliers, our sample includes relatively large customers of eligible firms, which are less likely to be financially constrained than the typical firm in the sample.⁷ in the Internet Appendix reports on the interconnections between suppliers and their customers by country pair.⁸

1.4 Empirical methodology

Our hypothesis is that eligible firms experience a decrease in cost of capital due to the CSPP and find it optimal to use the liquidity that they obtain through bond issuance to extend trade credit to marginal customers to which they were previously rationing trade credit. We thus view trade credit usage as driven by supply, consistent with growing evidence that customers can use the available slack in payment terms at little or no cost (Giannetti, Burkart, and Ellingsen 2011; Klapper, Laeven, and Rajan 2012; Murfin and Njoroge 2015; Barrot 2016; Breza and Liberman 2017; Giannetti, Serrano-Velarde, and Tarantino 2021).

Post is a dummy variable that takes the value of one in 2016, the year the CSPP is announced and implemented and thereafter, and zero otherwise. |$X_{i,t-1}$| is a set of firm-specific control variables that includes log(Assets), Cash, PPE, Net margin, and Liabilities. All control variables are lagged by 1 year. The regressions include firm fixed effects |$\eta_{i}$|⁠, industry-year fixed effects |$\eta_{j,t}$| (using the Fama-French 10-industry classification), and country-year fixed effects |$\eta_{c,t}$|⁠. Standard errors are clustered at the firm level to correct for heteroscedasticity and within-firm residual correlation.

A positive and significant estimate of the coefficient |$\beta_1 $| indicates that eligible firms extend more trade credit (Accounts receivable) or that customers of eligible firms receive more trade credit (Accounts payable) due to the CSPP, as long as demand and other shocks evolve similarly for treated and control firms. To validate this assumption, as is customary in the application of difference-in-differences methods, we evaluate whether there are any preexisting differential trends in the use of trade credit for eligible firms and their customers before the CSPP. This helps to address concerns that contemporaneous shocks might lead to a positive |$\beta_1 $| coefficient. This could include differential demand shocks to treated and control firms, or worsening credit conditions that might lead firms with better access to external finance to extend more trade credit during the implementation of the CSPP by the ECB. In both of these alternative hypotheses, we would expect to observe a gradual increase in the provision of trade credit before the start of the policy. The absence of preexisting trends suggests that the CSPP, and not differences in demand or in the ability to provide trade credit, lead eligible suppliers to increase trade credit provision.

To further validate our identifying assumption, we test whether the provision of trade credit increases around the investment-grade cutoff (i.e., the eligibility criteria) for eligible firms relative to noneligible firms after the CSPP announcement. As with the pre-trends tests, this helps to address the concern of whether firms of different size or with different access to external finance may have increased the provision of trade credit for reasons unrelated to the CSPP.

2. Effects of Unconventional Monetary Policy

In this section, we first examine whether eligible firms’ enhanced ability to tap debt markets, following the purchases of bonds by the ECB, is transmitted through trade credit in production networks. We then check the robustness of our results using different empirical approaches.

2.1 Accounts receivable

The CSPP decreased the cost of capital of eligible firms by decreasing bond yields and fostering bond issuance. This, in turn, allowed CSPP-eligible firms to extend more trade credit to marginal customers that were previously rationed.

To test our main hypothesis, we examine whether CSPP-eligible firms experienced an increase in the ratio of accounts receivable to sales (Accounts receivable) following the CSPP. Table 2 shows the results. The estimates in column 1 show that eligible firms (treatment group) experience a significant increase in the ratio of accounts receivable to sales of about 10 percentage points relative to noneligible firms (control group) after the announcement of the CSPP. The effect is economically significant as a 10-percentage-point increase in accounts receivable indicates that firms in the treatment group increase days receivable, relative to those in the control group, by about 36 days (⁠|$36 = 0.1 \times 360$|⁠), which is about one third of the sample average. The results remain robust when we add firm-specific controls (column 3) and country-year fixed effects (column 5) to the specification in column 1, which already includes firm fixed effects and industry-year fixed effects.

Columns 2, 4, and 6 show the evolution of the differences in accounts receivable between the treatment and control groups in the years before and after the CSPP (based on estimates in which Eligible is interacted with dummy variables for each event year). Figure IA.2 in the Internet Appendix shows the evolution of the differential effect in accounts receivable as estimated in column 2. We do not find statistically significant differences between treatment and control groups in the pretreatment period. While the increase in accounts receivable becomes statistically significant only after the start of the CSPP, we find a positive but insignificant increase in eligible firms’ accounts receivable already in 2015. A possible explanation is that the Public Sector Purchase Program (PSPP) implemented by the ECB in 2015 involved purchasing large amounts of government bonds and may have spilled over to investment-grade corporate bonds, which are often viewed as close substitutes (Krishnamurthy and Vissing-Jorgensen 2013). The trade credit provision by eligible firms accelerates after the implementation of the CSPP in 2016 and further increases in 2017, presumably after firms had time to issue bonds and then renegotiate trade credit terms with their customers.⁹ Consistent with these facts, we find that the interaction term |${\it Eligible} \times {\it 2015}$| coefficient is not statistically different from the |${\it Eligible} \times {\it 2016}$| coefficient (with a |$p$|-value between .5 and .6), but it is strongly statistically different from the |${\it Eligible} \times {\it 2017}$| coefficient (with a |$p$|-value between .04 and .08). Similarly, the sum of the interaction term coefficients for years 2014 and 2015 are statistically different from the sum of the interaction term coefficients for years 2016 and 2017 (with a |$p$|-value below .01 in all columns). Acceleration after the start of the CSPP makes it unlikely that unobserved demand shocks could drive our main results. We perform a variety of additional tests to support this conclusion.

First, panel A of Table 3 provides evidence supporting the spillover effects of the PSPP. If the PSPP drives the positive coefficient in 2015, we should observe a lessening of the spillover when we exclude firms rated AAA to A (i.e., the most highly rated firms) from the group of eligible firms in the estimation sample. This test has the advantage that the treated group includes the subset of eligible firms most affected by the ECB purchases during the implementation of the CSPP, that is, those rated A|$-$| to BBB|$ - $| (Grosse-Rueschkamp, Steffen, and Streitz 2019). Consistent with our conjecture, we find that the magnitude of the coefficient for eligible firms in 2015 is reduced to |$0.007-0.012$|⁠, or about a quarter of the magnitude of |$0.042-0.046$| when we include all eligible firms in Table 2.

Second, panel B of Table 3 exploits the discontinuity around the investment-grade cutoff. The sample in columns 1–3 includes only BBB|$+$| rated through BB|$-$| rated bonds, that is, bonds just above and below the investment-grade cutoff. The estimated coefficient for the Eligible dummy variable between 0.23 to 0.28 is nearly three times as large as the one in the full sample in Table 2. The larger effect is consistent with the evidence that firms just above the investment-grade cutoff experienced larger relative decreases in the cost of capital and issued more bonds than firms with higher ratings (Grosse-Rueschkamp, Steffen, and Streitz 2019). The larger estimated coefficient also suggests that unobserved heterogeneity for firms further away from the investment-grade cutoff tends to bias our estimates downward.

This interpretation is confirmed in columns 4–6, in which we use a wider bandwidth around the investment-grade threshold to A|$+$| rated through B|$-$| rated bonds. When we increase the heterogeneity of treated and control firms, the estimates for the coefficient of the Eligible dummy variable are lower than in columns 1–3 in Table 2, but still significantly larger than our baseline specifications in columns 1–3 in which all eligible firms are included in the treated group.

Third, we test the hypothesis that eligible firms that have issued bonds after the start of the CSPP, as opposed to eligible firms that did not issue bonds during this period, are responsible for the increase in accounts receivable. To perform this analysis, we decompose the Eligible dummy variable into two variables: the Eligible issuer dummy variable that takes the value of one if an eligible firm has issued bonds after the CSPP announcement date (i.e., between March 2016 and December 2017), and zero otherwise; and the Eligible nonissuer dummy variable that takes the value of one if an eligible firm has not issued bonds after the CSPP announcement date, and zero otherwise. Following the start of the CSPP, 70|$\%$| of the eligible firms issue bonds. Consistent with our interpretation of the evidence, panel C of Table 3 shows that the differential increase in accounts receivable is driven by eligible firms that have issued bonds after the start of the CSPP.

Fourth, we perform a placebo test by estimating the change in the ratio of accounts receivable to sales of U.S. investment-grade firms (the main criterion used by the ECB to determine eligible bonds), after the introduction of the CSPP.¹⁰ In this test, we use Compustat data, which only contains publicly listed firms, as Orbis offers a limited coverage of U.S. firms. Table IA.3 in the Internet Appendix reports the estimates. We find no evidence that U.S. investment-grade firms experience an increase in accounts receivable relative to control firms after the CSPP.

Finally, we explore to what extent the effects of the CSPP may spill over beyond eligible firms. Noneligible bonds may be affected by central banks’ bond purchases as investors rebalance their portfolios toward these bonds (and thus decreasing their yields). If the CSPP produces spillovers to noneligible bonds, our estimates are lower bounds for the actual effect of the CSPP. To check for this possibility, we estimate the accounts receivables regressions in Table 2, including the interaction of a dummy variable for firms whose bonds are not eligible under the CSPP (Noneligible) with the Post dummy variable. Table IA.4 in the Internet Appendix shows that the coefficient for the interaction term Noneligible |$\times $| Post is not statistically significant and the coefficient for the interaction term Eligible |$\times $| Post is similar to that in Table 2. We conclude that firms with noneligible bonds do not seem to extend more trade credit to their customers after the CSPP and are unlikely to attenuate our estimates of the trade credit channel of monetary policy.

2.2 Accounts payable

We also examine the effect of the CSPP on firms in the downstream network of eligible firms. We test whether customers of eligible firms are indeed the recipients of the increase in trade financing by eligible suppliers following the CSPP. The dependent variable is the ratio of accounts payable to sales (Accounts payable).

Table 4 shows the results. In panel A, the explanatory variable of interest is the interaction of the treatment group dummy variable Has eligible supplier with the Post dummy variable. The interaction term Has eligible supplier |$\times $| Post measures the differential effect on accounts payable between firms with eligible suppliers and those without eligible suppliers following the CSPP.

In column 1, we find that customers of eligible firms (treatment group) experience an increase in the ratio of accounts payable to sales of 4.8 percentage points relative to customers without a business relationship with eligible firms (control group). The effect is statistically and economically significant as customers of eligible firms relative to the control group benefit from an extension in payment terms of about 17 days (⁠|$17 = 0.048 \times 360$|⁠), which is about 20|$\%$| of the sample mean. The results remain robust when we add firm-specific controls (column 3) and country-year fixed effects (column 5) to the specification in column 1, which already includes firm fixed effects and industry-year fixed effect. Importantly, columns 2, 4, and 6 show that treatment and control groups follow parallel trends in the pretreatment period and that the increase in accounts payable is particularly pronounced in 2017. Taken together with the evidence in Tables 2 and 3 for suppliers’ provision of trade credit, the evidence on the timing of the increase in accounts payable is consistent with a causal effects of the CSPP. Similarly to the results for Accounts Receivable, the coefficient for the interaction term |${\it Eligible} \times {\it 2015}$| is not statistically different from the coefficient for |${\it Eligible} \times {\it 2016}$| (with a |$p$|-value between .4 and .8), but it is significantly different from the coefficient for |${\it Eligible} \times {\it 2017}$| (with a |$p$|-value below .01 in all columns). Panel A of Figure IA.3 in the Internet Appendix shows the evolution of the differential effect in accounts payable as estimated in column 2. This confirms that the increase in trade credit usage by customers of eligible firms is unlikely to be driven by demand shocks concomitant to the CSPP.

We also examine whether customers with a larger share of eligible suppliers (measured as a the number of eligible suppliers divided by the total number of suppliers) benefited more from the increase in trade credit by eligible firms. Panel B of Table 4 shows the results. The explanatory variable of interest is the interaction of the treatment variable Eligible Suppliers Share with the Post dummy variable.

In column 1, we find that the Eligible suppliers share |$\times $| Post coefficient is positive and statistically significant. A one-standard-deviation increase in the share of eligible suppliers (0.362 in the sample of customers of eligible suppliers) leads to an increase of 2.5 percentage points (⁠|$= 0.362 \times 0.069$|⁠) in accounts payable of customers of eligible suppliers (i.e., a delay in payment terms of about 9 days), which corresponds to 11|$\%$| of the sample mean. The results are robust across specifications in columns 3 and 5. In addition, columns 2, 4, and 6 show no evidence of significant preexisting differential trends between treatment and control groups indicating that the two groups of firms are unlikely to experience asymmetric demand shocks. Panel B of Figure IA.3 in the Internet Appendix shows the evolution of the effect in accounts payable, which confirms the inexistence of preexisting differential trends.

We also take into account the 82 customers of eligible firms whose bonds are eligible for purchase under the CSPP. To correct for the overlap between eligible firms and customers of eligible firms whose bonds are eligible for purchase under the CSPP, we estimate regressions in which the dependent variable is the ratio of accounts payable to sales (Accounts Payable) including both the Has eligible supplier |$\times $|Post and Eligible |$\times $|Post interaction terms. Table IA.5 of the Internet Appendix shows that the estimates of the Has eligible supplier |$\times $|Post coefficient are similar to those in Table 4.

It is also useful to ask whether following the announcement of the CSPP, eligible firms accelerate payments to their own suppliers, as there also could be positive spillovers in upstream industries (Barrot and Nanda 2020). Table IA.5 shows that eligible firms neither accelerate nor delay payments to their suppliers as the coefficient for Eligible |$\times $| Post is insignificant. This finding is consistent with the results of several papers showing that large customers receive plenty of trade credit from their smaller suppliers (e.g., Murfin and Njoroge 2015; Giannetti, Serrano-Velarde, and Tarantino 2021). It should not come at a surprise in our context because suppliers may be able to acquire new customers and expand their market share by extending trade credit (as we will show in Section 4.3), but eligible firms receive no obvious benefits for accelerating payments to smaller suppliers.

2.3 Robustness

A possible concern with our baseline results is that our firm-year panel regressions might be insufficient to cope with the heterogeneity of firms in the sample and the differential effects we estimate may capture asymmetric demand shocks rather than an increase in the supply of trade credit. To further establish the validity of our results, we employ several additional empirical approaches: (1) regressions with size-decile-by-year fixed effects; (2) regressions using a matched sample; (3) regressions excluding firms with less than €10 million in assets from the sample; (4) regressions with country-industry-year fixed effects; (5) regressions with industry-year fixed effects using two-digit SIC codes; (6) regressions with standard errors clustered at the industry-year level; and (7) regressions using the logarithm of the levels of trade finance.

First, we estimate specifications with firm size-decile-by-year fixed effects. We sort firms into size deciles each year based on total assets and interact each decile with yearly dummies. The size-by-year fixed effects control for time-varying demand shocks that could have asymmetric effects on firms of different sizes and contaminate our estimates of the trade credit channel.

Columns 1 and 3 of Table 5, panel A, show a statistically significant increase of about 10 percentage points in the accounts receivable of eligible firms relative to noneligible firms. Similarly, columns 1 and 3, panel B, we find a corresponding increase in the accounts payable of customers of eligible firms of about 3 percentage points relative to firms without a link to the upstream network of eligible firms. When we estimate the dynamic effects of the CSPP in columns 2 and 4 of panels A and B, we again find statistically significant effects for both accounts receivable of eligible firms and for accounts payable of their customers in the post-CSPP period. Overall, these findings mitigate the concern that asymmetric demand shocks to firms of different size drive our findings.¹¹