Do trade credits finance long-term investments?

Abstract

Our model indicates that European firms across all size categories use trade credits to purchase 6%–15% of tangible fixed assets and 14%–30% of intangible assets in the short-run. A long-run target adjustment model shows that large firms eventually replace most of this temporary trade credit financing with cheaper sources of funds. However, even in the long-run, small firms finance 4%–6% of tangible fixed asset purchases and 5%–10% of intangible fixed asset purchases using trade credit. Since smaller firms do not have the same access to bank credit as larger firms, trade credit is used to fund long-term investments Trade credit is therefore a key component in the capital structure decisions of smaller firms and it should be included in their weighted-average cost of capital calculations.

1. Introduction

Trade credit is the single most important source of short-term external finance in the United States constituting about 18% of total assets of US firms in the 1990s [1]. As reported by Barrot (2016), trade credit financing for US non-financial firms is about three times the size of bank loans [2]. It represents nearly one-third of the debt obligations for US small and medium sized enterprises (SMEs). McGuiness et al. (2018) report comparable results for Europe with accounts receivable comprising about 20% of total assets and trade credits representing 30% of all debt [3]. While recognizing the significance of trade credit as a part of working capital, the finance literature assumes that trade credits are only used to meet short-term financing needs and that firms do not finance long-term projects with trade credits [1]. The “conventional wisdom” as reflected in corporate finance textbooks is that trade credits do not enter the firm's capital structure decisions [4]. However, firms might find that trade credit is a quick and convenient source of funds, even for long-term investments. The purpose of this paper is to assess whether trade credits finance long-term investments in the short-run and whether this relationship extends into the long-run. We examine whether the “conventional wisdom” that trade credits are not part of the capital structure holds for a large sample of European firms.

The notion that trade credit serves purposes other than short-term financing and liquidity needs has received some attention in the recent finance literature. For example, firms facing financial distress may find trade credit to be a cheaper source of funds than bank loans [5]. During the Global Financial Crisis of 2007–2010, although financially unconstrained firms generally increased their reliance on bank loans, financially constrained SMEs increased their use of trade credit [6]. Some studies suggest that flexibility in the magnitude of trade credit extended is essential to SME survival during periods of financial crises [3,7], while younger and faster growing firms may be able to use trade credits to expand output and manage their rate of growth [8]. Finally, other research examining the use of working capital has shown that some US firms have used cash reserves to smooth out cash flow problems in the funding of long-term research and development expenditures [9] and that Chinese firms with higher levels of working capital over the period 2000–2007 were able to use these funds to reduce the impact of financing constraints on fixed investment [10].

In this paper we investigate whether European firms regularly purchase fixed assets, or fund long-term investments, using trade credits. Our results for 13.3 million European firm year observations over the years 1998–2017 indicate that firms of all sizes often use trade credit, especially in the short-run, to purchase both tangible and intangible fixed assets. The importance of trade credit in the firm's long-term capital structure varies across countries, industries, firm size, and listing status. Smaller firms with limited access to traditional bank lending channels are more likely to purchase fixed assets using trade credits than larger firms. Our suggested modification to the traditional capital structure is particularly important for SMEs, which Daskalakis et al. (2017) note are not merely scaled-down versions of large firms [11]. For smaller firms, trade credits should enter the weighted average cost of capital. In contrast, given the small economic significance of trade credits, the standard practice of omitting trade credits from the capital structure may be reasonable for larger firms.

2. Literature review

Petersen and Rajan (1997) state that “firms may be financed by their suppliers rather than by financial institutions” [1]. They discuss various theories regarding why firms either use or offer trade credits and then empirically estimate both supply and demand for trade credits in US firms for the year 1987. Larger US firms borrow more with trade credit than smaller firms, while firms extending trade credit to other firms are likely to be the larger, older, and more credit-worthy firms. They suggest that suppliers of trade credit may be better than banks at judging the credit worthiness of receiving firms, which explains the popularity of trade credit to finance the purchase of short-term assets.

Trade credit facilitates intra-firm business by reducing transactions costs, providing knowledge about the true state of a business, and they are a way to exercise some control over recipients [12]. It permits firms with cheaper credit to finance buyers who otherwise face expensive credit, and that it is a second party endorsement of product quality for firms receiving trade credit [13]. Receiving trade credit through accounts payable is a way to reduce imperfections in the financial market and perhaps in product markets. Over the period 1993–2009, European firms increased trade credit usage as a way of accelerating their growth rates [6].

Several recent studies have noted that the use of trade credit is more pervasive in some cultures than others. Extension of trade credit is more prevalent in countries with more collectivism, where there are greater differences between rich and poor customers, and in high-masculinity countries [14]. Firms in high-social capital countries use less trade credit than firms in low-social capital countries [15] and trade credit is more prominent in high-trust countries [16].

2.1. Trade credits and fixed assets

The average balance sheets of firms vary considerable across countries—particularly in the usage of fixed assets [17]. Over the period 1987–1991, tangible fixed assets constituted over one-half total assets for Canadian firms and less than one-fourth of total assets for French firms. Intangible fixed assets represented about 8% of total assets for US and French firms, but less than one percent of total assets for firms in Japan or the United Kingdom. Similarly, accounts receivable as a percentage of total assets ranged from 13% in Canada to 29% in France and Italy. Accounts payable varied from 11.5% of the liabilities plus equity in Germany, up to 17% for France.

Brown and Petersen (2011) is one of the first papers to point out that firms may use working capital for long-run purposes—in their case to maintain research and development expenditures from year to year [9]. Subsequently, Bartholdy et al. (2012) show that Portuguese SMEs during the 1990s funded about 36% of new investments in tangible long-term assets using short-term funding sources, such as short-term loans, leasing, and trade credits [18]. They classified trade credits as cheap or expensive depending upon how long the credits were outstanding and discovered that expensive trade credits provided up to 10% of the funding for long-run tangible assets. Also, working capital management is important for Chinese firms that have limited access to long-term capital markets [10]. Working capital acts as a buffer for fixed capital investment that shields the firm from changes in the availability of traditional long-term financing sources and it may finance fixed capital investments during periods of negative cash flows and external shocks [10]. Other studies have argued that short-term adjustments in the use of trade credits may help with firm survival during unfavorable economic conditions [3], 6], [7]. For SMEs, flexibility in the use of trade credit was crucial to survival during the financial crises from 2007 to 2011 [3,6,7,19].

2.2. Trade credits versus bank financing

Although not the primary focus of this paper, another critical issue in the empirical finance literature is whether trade credits and bank loans are complements or substitutes. Most studies find that trade credit and bank loans are substitutes and that firms use trade credit more extensively during financial crises. For example, financially unconstrained Spanish firms are more likely to use bank credit, while constrained firms rely more heavily upon trade credits [6]. Also, financially constrained firms use trade credits more than other firms and their trade credit usage rises as interest rates increase, or when bank credit becomes costlier [20]. However, trade credit usage during times of crisis does not always vary inversely with firm size. Casey and O'Toole (2014) report that the larger European SMEs (not the smaller firms) were more likely to turn to trade credit financing after financial crises of 2009–2011 [21]. Similarly, Chen et al. (2019) report that trade credits and bank loans were substitute financing sources for Chinese firms over the years 2002–2015 and that trade credit expands during financial crises and declines when standards for bank credit are relaxed [22].

Another strand of literature suggests that bank loans and trade credits are complements. This situation may arise because banks perform a credit assessment before extending loans and suppliers may view the receipt of a bank loan as a positive signal about firm quality that lessens the risk of further extending trade credit or increasing its maturity. This line of reasoning is consistent with the financial accelerator theory of Bernanke et al. (1996) which predicts a flight to quality during bad economic times [23]. This complementarity in financing relationship seems to amplify the concept of survival of the fittest during crises. To illustrate. Psillaki and Eleftheriou (2015) find a complementary relationship between bank loans and trade credits for French SMEs during the crisis years of 2007–2009 [24]. They report that the less profitable firms obtained less loans and less trade credits during the crisis, while the more profitable firms expanded the use of both financing sources during and after the financial crisis [24]. Such results indicate that governments concerned with helping SMEs should open more financing channels for small business during and after financial crises. In Europe, the firms historically most dependent on trade credit financing are often mid-size firms across a variety of industries that have been experiencing rapid sales growth [25]. To help some of these firms survive hard times, venture capital funds have been set up for young, high-tech firms in several European countries [26].

Finally, a third possibility is that the substitutability-complementarity relationship changes over time and that it may vary between countries. For example, Bussoli and Marino (2018) argue that trade credits are substitutes for bank loans for European SMEs during periods of tight credit, while they become complements during economic expansions [7].

3. Data

3.1. Definition of variables

Bureau van Dijk (BVD) provides accounting data for millions of private and public companies around the world. However, we limit our sample to Europe because the accounting data is more comprehensive than for other regions. Annual balance sheet data and the variable names, as defined by BVD are shown in Table 1 for the years 1999–2017. We also retrieved net income (NI), operating income (OPPI), age, and industrial code from the annual income statements.

Table 1.

Balance sheet items.

Assets		Liabilities
	Code		Code
Other Current Assets (cash)	OCAS	Trade credits	CRED
Debtors (Accounts receivable)	DEBT	Short term financial debt	LOAN
Inventory	STOK	Other current debt	OCLI
Current Assets (OCAS + DEBT + STOK)	CUAS	Current liabilities (LOAN + OCLI + CRED)	CULI
Tangible assets	TFAS	Long-term debt	LTDB
Intangible assets	IFAS	Other non-current liabilities	ONCL
Other fixed assets	OFAS	Non-current liabilities (LTDB + ONCL)	NCLI
Fixed assets (IFAS + TFAS + OFAS)	FIAS	Share holder funds (equity)	SHFD
Total assets (CUAS + FIAS)	TOAS

Definitions (from ORBIS online help facility).

OCAS: Other current assets (including cash but excluding STOK and DEBT).

DEBT: Trade account receivables (from clients and customers only).

STOK: Total inventories (raw materials + work in progress + finished goods).

CUAS: Total current assets.

TFAS: Tangible fixed assets, such as buildings and machinery.

IFAS: Intangible fixed assets, such as formation expenses, research expenses, goodwill, and development expenses.

OFAS: Other fixed assets, such shares and participation funds, and pension funds.

FIAS: Total fixed assets.

TOAS: Total assets.

CRED: Trade credit debts owed to suppliers and contractors.

LOAN: Short term financial debts (short-term bank loans + long-term financial debts payable within a year).

OCLI: Other current liabilities such as pension, personnel costs, taxes, intragroup debts, accounts received in advance.

LTDB: Long term financial debts (long-term bank loans and bonds).

ONCL: Other non-current liabilities (miscellaneous debt and pension loans + provisions + deferred taxes.

SHFD: Total equity (contributed capital + other shareholders funds).

From the initial sample of 190 million firm-year observations in the ORBIS historical file, we construct our sample through a series of data reducing steps. The final sample consists of about 13.3 million firm-year observations for 2.5 million European firms. Data examined include total assets (TOAS) and its component parts, where current assets (CUAS) = inventory (STOK) + Debtors or accounts receivable (DEBT) + Other Current Assets (OCAS), and long-term fixed assets (FIAS)= Intangible Fixed Assets (IFAS) + Tangible Fixed Assets (TFAS) + Other Fixed assets (OFAS). We required all assets to have nonnegative values with data present for years t and t-1 for FIAS and CUAS. Similarly, values for total shareholder funds (SHFD), non-current liabilities (NCLI), and current liabilities (CULI) = trade credit (CRED) + other current liabilities (OCLI) had to be available for years t and t-1. The component parts of NCLI = Long-Term Debt (LTDB) + Other Non-Current Liabilities (ONCL), and CULI = Short-Term Debts (LOAN) + Other Current Debt (OCLI) + Trade Credits (CRED) had to be nonnegative. After deleting records with missing observations, we trimmed the dataset by deleting 1% of the most extreme observations in tail of the distribution for each of the change variables used in the short run analysis (ΔCRED, ΔTFAS, ΔIFAS, ΔTOAS, ΔCUAS, ΔLTDB, ΔLOAN), as well as for the operating profit margin (OPM), Risk, and Growth variables. OPM is earnings before interest and taxes divided by revenue, Risk is measured by the Z-score, and Growth is the average increase in total assets over the past five years. Finally, we drop the financial service industry (two-digit NACE value = 25) and other service and semi-public firms (two-digit NACE value > 84) from the sample.

3.2. Data sample

At year end, we place each firm-year observation into a size group (Micro, Small, Medium, or Large) based on total assets using the size definitions adopted by the European Commission [27]. About 93% of the 13.3 million firm-year observations are for Micro and Small. Table 2 displays firm year observations by country and firm size. Italy has the largest number of firm-year observations (3.4 million), followed by France (3.2 million) and Spain (2.9 million). Table 3 lists the average ages of firms by size and country. The average age of firms across the size classes are the highest in Switzerland, Netherlands, and Great Britain and the lowest in Greece and Norway. Across countries, the average age of firms increases from 18 years for micro firms up to 28 years for large firms.

Table 2.

Firm-year observations by country and firm size (1999–2017).

The 21 countries are listed in alphabetical order based on abbreviations used by the European Commission.

Country	Number of firms	Avg. Number of obs. per firm	Number of firm year observations by firm size
			Micro	Small	Medium	Large	Sum
AT - Austria	1137	2.87	44	513	1228	1479	3264
BE - Belgium	279556	6.92	1622409	237233	52384	21682	1933708
CH - Switzerland	561	9.11	33	353	971	3754	5111
DE- Germany	24768	4.04	9559	30698	36016	23879	100152
DK - Denmark	8334	1.03	4824	2110	900	770	8604
ES - Spain	516791	5.69	2139687	602471	152787	48102	2943047
FI - Finland	45457	4.57	169818	25581	7591	4969	207959
FR - France	580940	5.55	2664364	418621	97779	41923	3222687
GB - Great Britain	56639	4.54	70403	90266	52193	44520	257382
GR - Greece	6003	3.62	11823	7390	1876	630	21719
IE - Ireland	1914	3.44	723	2598	1689	1583	6593
IS - Iceland	6392	3.73	20884	2260	461	266	23871
IT -Italy	664103	5.13	2196975	886010	254646	71277	3408908
LI - Liechtenstein	7	6.14		1	11	31	43
LT - Lithuania	1502	4.63	1158	3698	1509	587	6952
LU - Luxembourg	1036	2.44	589	512	612	818	2531
MT - Malta	544	3.06	725	505	307	125	1662
NL - Netherlands	1783	2.93	118	1143	1201	2769	5231
NO - Norway	96536	3.28	257150	43238	11236	5362	316986
PT - Portugal	181845	4.20	666591	73764	18176	5532	764063
SE - Sweden	5118	2.67	11899	1297	348	134	13678
Total			9,849,776	2,430,262	693,921	280,192	13,254,151

Table 3.

Average age of firms by size and country.

Age is calculated for each firm-year observation as the difference between the year of the data point and the year of incorporation. The 21 countries are listed in alphabetical order based on abbreviations used by the European Commission.

Country	Micro	Small	Medium	Large
AT - Austria	21.25	23.44	28.08	33.60
BE - Belgium	15.52	20.41	24.01	29.50
CH - Switzerland	50.02	47.63	52.25	47.04
DE- Germany	17.33	26.19	30.02	36.67
DK - Denmark	16.25	20.45	25.11	29.23
ES - Spain	13.85	17.12	19.90	23.08
FI - Finland	16.83	21.37	24.97	25.86
FR - France	14.18	21.07	24.60	28.06
GB - Great Britain	19.47	25.95	27.96	31.34
GR - Greece	8.87	9.39	11.31	14.76
IE - Ireland	17.45	25.53	26.63	19.99
IS - Iceland	13.05	19.08	18.87	21.32
IT -Italy	13.75	18.76	22.22	25.01
LI - Liechtenstein	–	12.00	10.50	38.38
LT - Lithuania	13.38	14.90	15.72	16.30
LU - Luxembourg	15.90	20.27	23.38	19.98
MT - Malta	18.12	21.46	22.49	20.82
NL - Netherlands	22.65	34.02	35.50	41.32
NO - Norway	12.91	14.72	14.94	15.41
PT - Portugal	16.16	21.56	25.31	26.68
SE - Sweden	18.05	24.22	32.21	36.36
Average for all countries	17.75	21.88	24.57	27.65

We next divide all balance sheet entries by total assets to obtain a common size balance sheet for each firm. Table 4 presents an average common size balance sheet for each size class. The aggregate totals for fixed assets, which represent long-term investment, range from 33% of assets for Micro firms up to 48% of assets for Large firms. Shareholder funds constitute 43% of funding for Micro firms and 39%–41% for other firms. Non-current liabilities (including long-term debt) make up 15% of funds for Micro firms and increase with firm size up to 21% for Large firms. Finally, trade credits comprise about 16–17% of funding for Small and Medium firms, but only 13% for Micro firms and 12% for Large firms.

Table 4.

Common size balance sheet by size class.

	Micro	Small	Medium	Large
Asset categories as a % of total assets
Current Assets
Other Current Assets (OCAS)	29.02	20.52	19.06	21.46
Debtors (DEBT)	22.46	24.21	23.48	17.38
Inventories (STOK)	15.66	18.79	18.56	13.46
Total Current Assets (CUAS)	67.13	63.52	61.11	52.30
Fixed Assets
Tangible Fixed Assets (TFAS)	23.95	26.47	25.62	25.47
Intangible Fixed Assets (IFAS)	4.32	2.37	2.31	3.87
Other Fixed Assets (OFAS)	4.59	7.64	10.96	18.36
Total Fixed Assets (FIAS)	32.87	36.48	38.89	47.70
Liability categories as a % of total liabilities plus shareholder funds
Short-Term Debt
Trade Credits (CRED)	13.34	16.57	16.26	12.40
Bank Loans (LOAN)	4.44	8.07	10.63	9.45
Other Current Liabilities (OCLI)	23.72	18.65	15.16	16.08
Total Current Liabilities (CULI)	41.50	43.28	42.05	37.93
Long-Term Liabilities
Long-Term Debt (LTDB)	10.37	11.28	12.03	14.16
Other Noncurrent Liabilities (ONCL)	4.69	6.31	5.98	7.26
Total Noncurrent Liabilities. (NCLI)	15.06	17.59	18.01	21.42
Shareholder Funds (SHFD)	43.44	39.12	39.94	40.65

4. Short-run analysis

Although the BVD database does not provide data on specific long-term investment projects, we can indirectly infer how firms finance the purchase of fixed assets by estimating annual changes in each of the major asset categories relative to changes in each funding source.1 For example, as shown in equations (1), (2), annual changes in long term investments are:

$$\Delta TFA{S}_{it}=\left(TFA{S}_{it}-TFA{S}_{i,t-1}\right)/TOA{S}_{it-1}$$ (1)

$$\Delta IFA{S}_{it}=\left(IFA{S}_{it}-IFA{S}_{i,t-1}\right)/TOA{S}_{it-1}\text{.}$$ (2)

The annual changes in the funding source variables ($\Delta CRE{D}_{it},\Delta LOA{N}_{it},\text{and}\Delta LTD{B}_{it}$) are similarly constructed and scaled by total assets, as is the annual change in current assets ($\Delta CUA{S}_{it}$). Our goal is to examine the impact of changes in assets and liabilities on the use of each funding source. To focus upon trade credit financing, we construct the following model that estimates the change in the annual use of trade credit caused by changes in assets, controlling for changes in other funding sources, periods of crisis, and firm profitability, as follows:

$$\Delta CRE{D}_{it}={\alpha }_0+{\alpha }_{1}Crise{s}_t+{\gamma }_1\Delta TFA{S}_{it}+{\gamma }_2\Delta IFA{S}_{it}+{\gamma }_1^c\Delta TFA{S}_{it}\times Crise{s}_t+{\gamma }_2^c\Delta IFA{S}_{it}\times Crise{s}_t+{\gamma }_3\Delta CUA{S}_{it}+{\gamma }_4\Delta LOA{N}_{it}+{\gamma }_5\Delta LTD{B}_{it}+{\gamma }_{6}OP{M}_{it}+{\epsilon }_{it}\text{.}$$ (3)

The variables in equation (3) represent changes from time t-1 to t, Crises is a dummy variable with the value of 1 for the years 2008–2011, and ε_it is an error term in the regression analysis.

The ${\alpha }_0$ coefficient represents the average annual change in trade credit financing over the period 1999–2017 (in absence of changes in other funding sources or in assets) and the ${\alpha }_1$ coefficient records whether this amount changes during financial crises. The coefficients γ₁ and γ₂ test whether trade credits finance long-term investments, while the coefficients ${\gamma }_1^c$ and ${\gamma }_2^c$ indicate whether this relationship changed during the financial and banking crises of 2008–2011. If the “conventional wisdom” about trade credit is correct, then all four of these coefficients should be insignificant. The γ₃ coefficient shows whether firms use trade credits to purchase current assets and we expect that γ₃ > 0 and highly significant.

Regarding the coefficient on the LOAN variable, short-term bank loans and trade credits are substitutes if γ ₃< 0 and complements if γ₃ > 0. Similarly, for the LTDB variable, trade credits and long-term debt are substitutes if γ₄ < 0 and complements if γ₄ > 0.

We now estimate equation (3) by ordinary least squares (OLS) regression with residuals double clustered over industry and years.2 As shown in Table 5, the γ₁ and γ₂ coefficients for the ΔTFAS and ΔIFAS variables are positive and significant. The interpretation of the γ₁ coefficient is that Small firms finance 6% of tangible fixed asset purchases with trade credit. Micro and Medium size firms finance 12–13% of new tangible asset purchases using trade credit, while Small and Large firms use trade credit to finance about 15% of tangible asset purchases. The financial crises of 2008–2011 had negligible impact upon whether Micro or Large firms used trade credits to purchase long-term tangible assets. However, Small and Medium size firms decreased their use of trade credit to finance tangible fixed assets during these crisis years.

Table 5.

Short-run analysis.

variable	Micro	Small	Medium	Large
Constant (α0)	0.0022	0.0071	0.0091	0.0072
t-statistic	(1.67)	(2.84)	(3.67)	(2.75)
Crisest	0.0057	0.0001	−0.0043	−0.0047
t-statistic	(1.36)	(0.03)	(-1.11)	(-1.02)
ΔTFASit	0.1310	0.0643	0.1232	0.1483
t-statistic	(8.41)	(4.03)	(4.90)	(2.54)
ΔIFASit	0.1449	0.1378	0.2989	0.2274
t-statistic	(5.88)	(4.81)	(5.65)	(2.89)
ΔTFASitx Crisest	0.0122	−0.0310	−0.0944	−0.0698
t-statistic	(1.88)	(-2.21)	(-4.37)	(-1.52)
ΔIFASitxCrisest	−0.1928	0.0007	0.0189	0.1127
t-statistic	(-1.13)	(0.01)	(0.45)	(0.56)
ΔCUASit	0.2293	0.1867	0.0916	0.0028
t-statistic	(8.43)	(6.30)	(3.37)	(1.52)
ΔLOANit	0.0200	0.1156	0.1220	0.1743
t-statistic	(0.62)	(3.88)	(6.48)	(3.05)
ΔLTDBit	−0.0997	−0.0467	−0.0345	0.0087
t-statistic	(-5.41)	(-1.94)	(-3.13)	(1.05)
OPMit	−0.1135	−0.0355	0.0010	0.0225
t-statistic	(-7.41)	(-1.54)	(0.05)	(2.38)
N	9,674,333	2,399,468	684,430	277,090
R-Squared	0.21	0.21	0.18	0.07

For intangible fixed assets, the results are more striking than for tangible fixed assets. Trade credits finance about 14% of intangible asset purchases for Micro and Small firms, 30% for Medium firms, and 23% for Large firms. The negative ${\gamma }_2^c$ coefficient is statistically insignificant for all size groups, meaning that financial crises had no impact on the use of trade credit financing to buy intangible assets.

Regarding the other variables in Table 5, a one Euro increase in current assets increases the use of trade credits by 0.09–0.23 Euros for Micro, Small and Medium sized firms, but only by 0.003 Euros for Large firms. Also, an increase in earnings (cash flows), as measured by the γ₆ coefficient on the OPM variable, decreases the use of trade credits for Micro firms. However, the coefficient is insignificant for Medium and Small firms, and positive for Large firms.

The main conclusion from our short-run analysis is that European firms use trade credits to purchase both tangible and intangible fixed assets. Our results are contrary to the “conventional wisdom” that short-run financing sources are for current assets and that long-run financing is for long-lived assets. We also find that short-term debt and trade credits are complementary financing sources across all sizes of firms, while long-term debt and trade credits are substitutes for Micro, Small, and Medium-sized firms and that the relationship is insignificant for Large firms.

5. Long-run analysis

5.1. Long-run model

To assess whether the statistical significance of the trade credit variables in Table 5 are a temporary phenomenon, we add trade credits into a capital structure target adjustment model developed by Flannery and Rangan (2006) and later used in other studies of capital structure [[28], [29], [30]]. We assume that the optimal level of trade credit for firm i at time t ($CRE{D}_{it}^{*}$) in equation (4) depends upon historical values of balance sheet variables (X_it-1) from the previous year, plus past and current income statement variables (Z_it) as follows:

$$CRE{D}_{it}^{*}=\lambda X_{it-1}+\beta Z_{it}+\alpha F{E}_{it}\text{,}$$ (4)

where λ, β, and ɑ are regression parameters and FE_it is a fixed effect operator that captures unobserved effects not contained in the X_it-1 and Z_it variables. If firms adjust trade credit financing each year in accordance with a target adjustment model, the annual change (ΔCRED_it) is:

$$\Delta CRE{D}_{it}=\phi \left(CRE{D}_{it}^{*}-CRE{D}_{it-1}\right)\text{.}$$ (5)

Annual adjustments depend upon the direction and magnitude of the deviation of the previous period level of trade credits (CRED_it-1) away from the desired long-run level of trade credits ($CRE{D}_{it}^{*}$), and $\phi$ represents the speed of adjustment. Rearranging equation (5) gives an alternative representation of the model in equation (6):

$$CRE{D}_{it}=\phi CRE{D}_{it}^{*}+\left(1-\phi \right)CRE{D}_{it-1}\text{.}$$ (6)

Substituting the financial variables used in the previous section into equation (4), the desired long-run level of trade credits in equation (7) is:

$$CRE{D}_{it}^{*}=\omega +{\phi }_{1}TFA{S}_{it-1}+{\phi }_{2}IFA{S}_{it-1}+{\phi }_{3}CUA{S}_{it}+{\phi }_{4}LOA{N}_{it}+{\phi }_{5}LTD{B}_{it}+{\phi }_{6}Ris{k}_{it-1}+{\phi }_{7}Growt{h}_{it-1}+{\phi }_{8}Ag{e}_{it}+{\phi }_{9}OP{M}_{it}+\alpha F{E}_{it}+{\epsilon }_{it}\text{.}$$ (7)

Whether the right-hand side variables in equation (7) are current or lagged one period, depends on the information available to the firm at the time of the trade credit financing decision. LOAN and LTDB are current period variables because the firm can look at its bank statements to see the current amount of short-term and long-term credit. Similarly, the firm manages its current assets and liabilities on a short-term basis, so CUAS is a current period variable. Age is a readily discernible current period variable, while OPM is a proxy for cash flow, which is a figure currently available to the firm. For small firms, tangible and intangible fixed assets might be calculated only once a year for the annual report. In our analysis, we lag TFAS and IFAS one period, as well as the Risk and Growth variables because they depend upon historical data.

Substitution of the right-hand side variables from equation (7) into the target adjustment model yields equation (8):

$$CRE{D}_{it}=\phi \omega +\phi {\phi }_{1}TFA{S}_{it-1}+\phi {\phi }_{2}IFA{S}_{it-1}+\phi {\phi }_{3}CUA{S}_{it}+\phi {\phi }_{4}LOA{N}_{it}+\phi {\phi }_{5}LTD{B}_{it}+\phi {\phi }_{6}Ris{k}_{it-1}+\phi {\phi }_{7}Growt{h}_{it-1}+\phi {\phi }_{8}Ag{e}_{it}+\phi {\phi }_{9}OP{M}_{it}+\alpha F{E}_{it}+\left(1-\phi \right)CRE{D}_{t-1}\text{.}$$ (8)

After notational simplification, the estimation model is:

$$CRE{D}_{it}={\lambda }_{1}TFA{S}_{it-1}+{\lambda }_{2}IFA{S}_{it-1}+{\lambda }_{3}CUA{S}_{it}+{\lambda }_{4}LOA{N}_{it}+{\lambda }_{5}LTD{B}_{it}+{\beta }_{1}Ris{k}_{it-1}+{\beta }_{2}Growt{h}_{it-1}+{\beta }_{3}Ag{e}_{it}+{\beta }_{4}OP{M}_{it}+{\alpha }_{it}F{E}_{it}+\delta CRE{D}_{it-1}+{\epsilon }_{it}\text{.}$$ (9)

As previously defined, CRED are trade credits, TFAS are tangible assets, IFAS are intangible assets, CUAS are current assets, LOAN is short-term debt, and LTDB is long-term debt, and we scale each of these variables by total assets in the previous period. λ₁ - λ₅ are regression parameters for the five balance sheet variables, β₁ – β₄ are regression parameters for the four income statement control variables (Risk, Growth, Age, and OPM). FE_it is the fixed effects operator for individual firms in each year and it acts as constant term for each firm in each year. CRED_it-1 are trade credits in the previous year, ɑ_it and δ are regression coefficients, and ε_it is the error term. We assume that the firm chooses its investments first and then decides its financing. Thus, TFAS, IFAS, and CUAS are exogenous, as are the Growth, Age, Risk, and OPM control variables. However, LOAN and LTDB are determined endogenously along with the annual level of trade credit.

λ₁ and λ₂ are the primary coefficients of interest in our analysis and in equation (9) these coefficients test for independence between long-run investments and trade credits. If these coefficients are zero, then firms do not use trade credits to purchase fixed assets and the traditional story that trade credits are NOT a part of the firm's capital structure is correct. However, if either λ₁ > 0 or λ₂ > 0 and statistically significant, then trade credits are a part of a firm's weighed average cost of capital. As in the short-run analysis, LOAN and CRED are long-run substitutes if λ₄ < 0 and complements if λ₄ > 0. If λ₅ < 0, long-term debt and trade credits are substitutes, and λ₅ > 0 indicates complementarity.

5.2. Long-run estimation results

Although equation (9) can be estimated by OLS with fixed effects, correlation between the fixed effects operator and the lagged dependent variable can lead to biased coefficient estimates [30]. The bias is inversely related to the number of time periods (T) for each included firm and the fixed effects estimator can be severely biased, even in samples as large as T = 30 [31]. OLS estimation is also problematic due to endogeneity of the LOAN and LTDB variables. Since our data sample has an average of T = 5 among the four firm sizes, we present results in Table 6 using the Blundell-Bond panel estimator [32]. This technique handles the endogeneity of the LOAN and LTDB variables and reduces the small sample bias problem inherent in OLS fixed effects estimation.3 Also, the Blundell-Bond estimator can deal with the endogeneity of the LOAN and LTDB variables and the literature suggests that the estimator is well-suited for small T, large N panels, such as our data set [33].

Table 6.

Long-run analysis.

	Micro	Small	Medium	Large
TFASit-1	0.0435	0.0586	0.0552	0.0127
p-value	(0.04)	(0.05)	(0.05)	(0.02)
IFASit-1	0.0954	0.0875	0.0495	0.0140
p-value	(0.04)	(0.04)	(0.04)	(0.04)
CUASit	0.1202	0.1566	0.1364	0.0823
p-value	(0.02)	(0.04)	(0.04)	(0.02)
LOANit	0.1965	0.1031	0.0347	−0.0259
p-value	(0.36)	(0.73)	(0.10)	(0.03)
LTDBit	−0.0010	−0.0411	−0.0367	−0.0129
p-value	(0.20)	(0.13)	(0.04)	(0.02)
OPMit	−0.0959	−0.0859	−0.0497	−0.0245
p-value	(0.01)	(0.12)	(0.03)	(0.01)
Riskit-1	0.000027	0.0001	0.00001	0.00001
p-value	(0.29)	(0.06)	(0.04)	(0.11)
Growthit-1	−0.0420	−0.0134	−0.0075	−0.0010
p-value	(0.97)	(0.34)	(0.63)	(0.06)
Ageit	−0.0391	−0.0339	−0.0304	−0.0078
p-value	(0.13)	(0.03)	(0.06)	(0.07)
Credit-1	0.4816	0.5692	0.5232	0.5934
p-value	(0.0)	(0.0)	(0.0)	(0.0)
N	7263746	1786993	505737	212116

Boldfacing indicates significance at the 5% level.

Since it is unlikely that residuals in equation (9) are independent, clustering of residuals improves the accuracy of the t-statistics for the regression coefficients. Also, since two-way clustering of residuals by individual firm and year led to estimation problems, we employed three-way clustering of residuals by country, industry, and year. Results in Table 6 for the four size groups are based on Blundell-Bond estimates of a model that includes firm and year fixed effects, with residuals bootstrapped and clustered by country, industry, and year [34].

For Micro, Small and Medium size firms, the λ₁ coefficient on TFAS indicates that in the long-run that 4–6% of tangible asset purchases are finance by trade credits, but Large firms finance only 1.3% of tangible assets with trade credits. Since λ₂ > λ₁ across all sizes of firms, European firms are more likely to purchase intangible assets, rather than tangible assets, using trade credits. Micro firms finance 9.5% of new intangible assets with trade credit, Small firms fund 8.75%, and Medium-sized firms fund 4.95%. Large firms, in the long-run, only fund 1.40% of intangible asset purchases with trade credit. Nevertheless, our results indicate that firms purchase fixed assets using trade credits, even in the long-run. For large, and to a lesser extend for Medium-sized firms, the standard procedure of not including trade credits in the weighted average cost of capital (WACC) may be appropriate. However, for Micro and Small where trade credits finance 9–10% of intangible fixed assets and 4–6% of tangible asset purchases, the standard practice of ignoring trade credits in calculating WACC is questionable.

The long-run λ₁ and λ₂ coefficients for the TFAS and IFAS variables in Table 6 are smaller than their short-run counterparts (γ₁ and γ₂) in Table 5. In the short-run, the various sized firms purchased 6%–15% of TFAS and 14%–30% IFAS using trade credits. Such differences might occur if firms initially use trade credits to purchase fixed assets and then search for cheaper sources of financing. This is particularly noticeable for Large firms that fund 15% of TFAS with trade credits in the short-run, but only 1.3% in the in the long-run.

The λ₄ coefficient for the LOAN variable is negative and significant for Large firms. The two funding sources are long-run substitutes for Large firms, but this coefficient is insignificant for Micro, Small, and Medium firms. Regarding long-term debt, the λ₅ coefficient is significantly negative for Medium and Large firms, suggesting complementarity between trade credit and long-term debt. However, this coefficient is insignificantly negative for Small and Micro firms.

Except for the Medium firms, OPM is significantly negative indicating a long-run negative relationship between trade credit financing and profitability. Over time, firms may be moving toward their preferred capital structure by using internally generated funds to replace expensive trade credits.

For Micro, Small, and Medium firms, the lack of significance for the Growth coefficient appears to suggest that trade credits and asset growth are unrelated. However, because the growth variable requires five years of data, it may not show the situation for the youngest and fastest growing Micro firms, and we examine this possibility in the next section. The Risk coefficients are positive for all size groups, but only significant for Medium size firms--suggesting only a mild relationship between firm risk and its reliance on trade credit. Finally, the negative Age coefficients indicate that older firms of all sizes rely less on trade credits than their newer counterparts—possibly because these firms have better access to bank financing and less need to rely on expensive trade credits.

The results above suggest the following interpretation for trade credits. For Micro firms, trade credits finance fixed asset purchases in both the short and long-run. The level of trade credits is lower if firms can obtain long term debt, so trade credits and long-term debt are short-run substitutes. Over time, the level of trade credit financing for tangible fixed assets drops from 13% to 4%. The level of external debt does not increase, but internal funds replace trade credit financing, as confirmed by the negative OPM coefficient.

Small firms finance 6% of tangible fixed assets by trade credits in the short run and the sign on LOAN is positive. Thus, trade credits and short-term debt are short-run complements for Small firms. The share of trade credit financing for tangible assets remains at 6% in the long-run and neither short-term loans nor long-term debt influence the long-run level of trade credits for Small firms.

Medium firms use trade credits to finance 12% of investments in tangible assets in the short-run, but trade credit and long-term debt are long-run substitutes versus short-run complements. In the long-run, the share of financing of tangible assets by trade credits falls to 5% for Medium firms as they replace trade credit with internal funds (both LTDB and OPM are negative and significant).

Large firms initially finance 15% of tangible assets and 23% of intangible assets with trade credits. In the long-run, the trade credit share of financing drops to 1.3% for tangible assets and to 1.4% for intangible assets as firms switch toward greater reliance on short-term debt, long-term debt, and internal funds.

In terms of calculating WACC, the textbook method may be reasonable for Large firms. The long-run share of trade credit financing is only 1.3%–1.4% of tangible and intangible fixed assets, respectively. For Micro, Small, and Medium firms the trade credit share of financing is 4%–6% for tangible assets and 5%–10% for intangible assets. As shown in Table 3, intangible assets only constitute about 2%–4% of all assets of the three smaller size classes of firms. Although the trade credit financing share is high for intangible assets, intangible assets are only a small proportion of total assets. However, given the long-run importance of trade credit financing of tangible fixed asset purchases for Micro, Small, and Medium firms, trade credit should be included as a component of capital structure. For SMEs, trade credits should be a part of the WACC, whenever feasible to include.

6. Robustness tests

6.1. Impact of financial crises

The banking and financial crises of 2008–2011 reduced the amount of credit obtainable from banks to finance new investments and previous research has suggested that firms increased their use of trade credits to get through these crisis years [6,21]. We can assess whether firms also increase their use of trade credits to finance long-term investments during financial crises.

We augment the target adjustment model of equation (9) to examine the long-run impact of financial crises on the use of trade credits as follows4:

$$CRE{D}_{it}={\alpha }_{1}Crisi{s}_t+{\lambda }_{1}TFA{S}_{it-1}+{\lambda }_1^{c}TFA{S}_{it-1}\times Crise{s}_t+{\lambda }_{2}IFA{S}_{it-1}+{\lambda }_2^{c}IFA{S}_{it-1}\times Crise{s}_t+{\lambda }_{3}CUA{S}_{it}+{\lambda }_{4}LOA{N}_{it}+{\lambda }_{5}LTD{B}_{it}+{\beta }_{1}Ris{k}_{it-1}+{\beta }_{2}Growt{h}_{it-1}+{\beta }_{3}Ag{e}_{it}+{\beta }_{4}OP{M}_{it}+{\alpha }_{it}F{E}_{it}+\delta CRE{D}_{it-1}+{\epsilon }_{it}\text{.}$$ (10)

As in Table 5, we estimate this model using the Blundell-Bond estimator with LOAN and LTDB as endogenous variables. The bootstrapped p-values are clustered over country, industry, and year. In Table 7, the dummy variable Crises = 1 for the years 2008–2011, and Crises = 0 in other years. Like results in Table 6, the λ₁ and λ₂ coefficients for TFAS and IFAS are positive and statistically significant in most instances. Although firms use trade credits to finance long-term investments outside crisis years, the insignificance of the α₁ coefficient indicates that overall usage of trade credits does not increase during financial was not much different during crises years than in other years. The ${\lambda }_1^c$ coefficient is positive and statistically significant only for Small firms, and insignificant for Micro and Large firms (it was not possible to bootstrap the p-values for the Medium firms). Only Small firms increased the use of trade credit to purchase tangible fixed assets during financial crises and the financing share only increased from 2.3% to 3.4%. A possible explanation is that creditors viewed Micro firms as too risky during the crises years and were not willing to extend additional trade credit, while Medium and Large firms had less need to increase their use of trade credit. The ${\lambda }_2^c$ coefficient is positive for all but Large sized firms, but only statistically significant at the 5% level or higher for Large firms. The decrease in trade credit financing of intangible assets during the crises years is economically insignificant (but statistically significant). Overall, financial crises had only a limited impact on the use of trade credits for financing fixed assets. Most European firms did not increase trade credit usage during the financial crises of 2008–2011, nor did they substitute trade credits for bank financing to survive the crises years. In contrast, recent evidence on the use of trade credit financing for listed Chinese firms for the years 2017–2022, indicates that trade credit financing increased for these large corporations during and after the COVID-19 pandemic [35]. Even large Chinese corporations have been impacted by this crisis and the role of trade credit financing has increased in recent years.

Table 7.

Impact of financial and banking crises.

variable	Micro	Small	Medium	Large
Crisest	.0081	−.0031	−.0067	−.0041
p-value	(0.277)	(0.275)	(0.259)	(0.451)
TFASit-1	.0553	0.0233	.0529	.0108
p-value	(0.031)	(0.063)	(0.051)	(0.026)
TFASit-1 x Crisest	.0017	.0107	.0093	.0051
p-value	(0.911)	(0.039)	n.c.	(0.075)
IFASit-1	.0949	.0766	.0458	.0143
p-value	(0.101)	(0.043)	(0.032)	(0.043)
IFASit-1 x Crisest	.0030	.0339	.0003	−.0006
p-value	(0.839)	(0.110)	(0.547)	(0.046)
N	7,263,746	1,786,993	505,737	212,116

n.c. means not computed, because the multiway-clustered variance estimate was not positive.

Boldfacing indicates significance at the 5% level.

6.2. Robustness tests for specific types of firms

Our general results across all types of European firms do not support the Carbó-Valverde et al. contention that firms expand the use of trade credit during financial crises to lessen the impact of tighter bank credit [6]. However, trade credit usage could vary across firms by characteristics other than firm size. To address this possibility, we expand equation (10) by including an Xvar dummy variable, as follows:

$$CRE{D}_{it}={\alpha }_{1}Crise{s}_t+{\alpha }_{2}Xva{r}_{it}+{\alpha }_{3}Xva{r}_{it}\times Crise{s}_t+{\lambda }_{1}TFA{S}_{it-1}+{\lambda }_{2}IFA{S}_{it-1}+{\lambda }_1^{c}TFA{S}_{it-1}\times Crise{s}_t+{\lambda }_2^{c}IFA{S}_{it-1}\times Crise{s}_t+{\theta }_{1}TFA{S}_{it-1}\times Xva{r}_{it}+{\theta }_{2}IFA{S}_{it-1}\times Xva{r}_{it}+{\theta }_1^{c}TFA{S}_{it-1}\times Xva{r}_{it}\times Crise{s}_t+{\theta }_2^{c}IFA{S}_{it-1}\times Xva{r}_{it}\times Crises+{\lambda }_{3}CUA{S}_{it}+{\lambda }_{4}LOA{N}_{it}+{\lambda }_{5}LTD{B}_{it}+{\beta }_{1}Ris{k}_{it-1}+{\beta }_{2}Growt{h}_{it-1}+{\beta }_{3}Ag{e}_{it}+{\beta }_{4}OP{M}_{it}+{\alpha }_{it}F{E}_{it}+\delta CRE{D}_{it-1}+{\epsilon }_{it}\text{.}$$ (11)

The Xvar dummy variable above takes on four different values in each of the four regressions presented in Table 7. The fixed effect operator (FE_it) reflects the constant terms across firms and years. As a robustness check, we examine the importance of trade credit financing for (1) high-growth firms, (2) high-risk firms, (3) young firms, and (4) financially constrained firms relative to other firms. The sample for this exercise is Micro and Small firms (excluding Medium and Large firms). We employ the Blundell-Bond estimator with fixed effects and triple clustered residuals to estimate each of the four regression models for Xvar.

6.2.1. High-growth firms

High-growth firms require considerable funding to finance growth, but these firms are often associated with elevated levels of asymmetric information and agency problems that reduce the likelihood of obtaining bank credit. Such firms may rely more heavily on trade credit finance than other firms. Although the Growth variable was insignificant in equation (9) across all firms in Table 5, we now focus on the first Xvar variable. It applies to high-growth versus low-growth firms. For each year, Micro and Small firms are sorted based on the 5-year percentage growth rate of a firm's total assets. The fastest growing quartile of firms is assigned a value of Hgrowth = 1, the lowest quartile or slowest growing firms are assigned a value of Hgrowth = 0, and the remaining 50% of firms in each year are dropped when estimating equation (11). The model is written in deviation firm, so the dummy variable coefficients presented in Table 7 represent the differences between high and low-growth firms. When estimating equation (11) for Xvar = Hgrowth, we drop the Growth variable.

As shown by the α₂ coefficient for the Xvar variable, high-growth firms use more trade credits than low-growth firms. During crises, low-growth firms do not expand the usage of trade credits (as shown by the insignificance of α₁ coefficient for the Crises dummy), but high-growth firms increase the use of trade credits (since α₂ + α₃ is positive). The λ₁ and λ₂ coefficients for TFAS and IFAS are positive and statistically significant for TFAS, low-growth Micro and Small firms finance 4% of tangible fixed investments by trade credits. However, as shown by the ${\lambda }_1^c$ and ${\lambda }_2^c$ coefficients, the use of trade credit financing by low-growth firms to purchase long-term assets is the same inside and outside the financial crises. The θ₁ and θ₂ coefficients for the TFASxXvar and IFASxXvar variables show that high-growth firms have the same usage of trade credits as low growth firms since the coefficients are insignificant. During crisis years, the dependence on trade credits for financing fixed assets does not change (${\theta }_1^c$ and ${\theta }_2^c$ are and insignificant). These results show that high growth firms depend more upon trade credit financing than low growth firms and that dependence increases during the financial crises. However, financial crises do not influence the use of trade credits for financing fixed assets for either high or low growth firms.

6.2.2. High-risk firms

High-risk firms face a similar asymmetric information problem as high-growth firms and suppliers may be less willing to extend trade credit to risky firms. Xvar is labeled Hrisk, where risk is based on the standard deviation of three to five-year earnings (5 years, whenever available). Following the same procedure as for high-growth firms, we sort and rank firms by risk. Hrisk = 1 for the riskiest 25% of Micro and Small firms and Hrisk = 0 for the 25% of firms with the lowest earnings risk. We discard the middle 50% of the sample and drop the Risk variable when estimating (11) for the Xvar=Hrisk variable.

The Xvar dummy is negative and significant. Thus, high-risk firms use less trade credits compared with low-risk firms outside the crises. None of the other variables are significant, thus there is no difference between high and low-risk firms during the crises nor with respect to financing fixed assets.

6.2.3. Young firms

We define the dummy variable Young = 1 for a firm that is under seven years old. Such firms may have difficulties in obtaining bank credit and are likely to turn to trade credit financing. For all other firms, Young = 0. Our sample for this regression again consists of all Small and Micro firms. The positive and significant α₂ coefficient for Xvar indicates that younger firms have more trade credits than older firms. Although the coefficient is statistically significant, the level of trade credits for young firms is only 0.6% higher than for older firms. Regarding the TFAS and IFAS variables, older firms use trade credits to purchase tangible fixed assets. Younger firms also purchase fixed assets with trade credits, but to a lesser extent than older firms (θ₁ for the TFASxYoung coefficient is negative and significant), and again the economic effects is small. Financial crises do not appear to impact trade credit usage, again contradicting the results of Carbó-Valverde et al. [6].

6.2.4. Financially constrained firms

Our data set does not identify which firms are financially constrained, but following Bartholdy et al. [18], we define financially constrained firms as those firms that have overdrawn their trade credit terms. Ideally, we would like to have detailed information about the terms and length of trade credit contracts for each firm, but such information is not available. Instead, we estimate a standard length of contract for each industry and the number of credit days for each firm in the industry as follows:

$$\text{Credit}\text{days}=\frac{\text{trade}\text{credits}}{\text{cost}\text{of}\text{goods}\text{sold}/365}\text{.}$$ (12)

Although data on cost of goods sold in equation (12) are only available for 275,626 of the 13.3 million observations in the initial sample, this procedure can yield insight into differences in trade credit usage between financially constrained firms and other firms. We define firms to be financially constrained if the number of credit days exceeds the industry average by more than 1.96 standard deviations [18]. To illustrate, across all firms and years, the average number of days in the trade credit contract is 37 days and the standard deviation is 19 (rounded to an integer). The actual cut-off for the constrained classification varies by industry and year, but on average, firms with more than seventy-six credit days are financially constrained.

The Crises dummy in the Constrained regression, as shown in Table 8, is insignificant. However, as expected, the Xvar dummy is positive, large, and strongly significant. Such findings are consistent with the literature [6,20,21] that financially constrained firms rely more heavily upon trade credits than other firms. The TFAS and IFAS variables indicate that both non-constrained and financially constrained firms use trade credits to finance investments during normal times, but trade credit usage declines during financial crises. However, the TFASxXvar and IFASxXvar dummies suggest that financially constrained firms use significantly less trade credits compared with non-constrained firms in purchasing fixed assets. This relationship is unchanged during financial crises. Although financially constrained firms use more trade credits than non-constrained firms, they are less likely than non-constrained firms to use trade credits to purchase tangible or intangible fixed assets. An explanation may be that financially constrained firms use trade credits for short-term survival, rather than for long-term investments. Alternatively, suppliers may not be willing to finance long-term investments for financially constrained firms.

Table 8.

Robustness checks.

Variable	Hgrowth	Hrisk	Young	Constrained
Xvarit	.0345	−0.076	−0.0401	0.0814
p-value	(0.029)	(0.047)	(0.026)	(0.011)
Crisest	.0036	0.0132	0.0064	−0.0008
p-value	(0.238)	(0.780)	(0.497)	(0.937)
Crisest x Xvarit	0.0037	−0.0115	−0.0010	0.0032
p-value	(0.024)	(0.993)	(0.162)	(0.976)
TFASit-1	.0424	0.0285	0.0479	0.0745
p-value	(0.027)	(0.051)	(0.029)	(0.022)
IFASit-1	.0929	0.1193	0.1056	0.0719
p-value	(0.070)	(0.117)	(0.089)	(0.027)
Crisest x TFASit-1	.0030	−0.0037	0.0039	0.0010
p-value	(0.916)	(0.926)	(0.806)	(0.263)
Crisest x IFASit-1	0.0032	0.0145	0.0107	0.0004
p-value	(0.616)	(0.813)	(0.550)	(0.471)
TFASit-1 x Xvarit	.0042	0.0270	−0.0073	−0.0543
p-value	(0.076)	(0.168)	(0.015)	(0.010)
IFASit-1 x Xvarit	−.0259	−0.0392	−0.0087	−0.0980
p-value	(0.111)	(0.415)	(0.433)	(0.014)
TFASit-1 x Crisest x Xvarit	−.0018	0.00947	0.0020	−0.0021
p-value	(n.c.)	(0.886)	(0.889)	(0.909)
IFASit-1 x Crisest x Xvarit	.0784	0.0014	0.0039	−0.0115
p-value	(0.297)	(n.c.)	(0.898)	(0.682)
N	4,596,998	4,664,569	9,311,022	150,111

n.c. means not computed, because the multiway-clustered variance estimate was not positive.

Boldfacing indicates significance at the 5% level.

In summary, financially constrained firms (likely to be young, high growth and high-risk firms) use more trade credits and are more likely to use expensive trade credits than non-constrained firms (likely to be older, lower-growth, and lower-risk firms). Financially constrained firms use less trade credits to purchase fixed assets than unconstrained firms, but this outcome may occur because suppliers are careful in extending trade credit to financially constrained firms. Again, in contrast to results previously reported in the literature, the use of trade credit is unchanged during financial crises.

7. Summary and conclusions

Traditional finance theory assumes that firms do not use short-term financing sources, such as trade credits, to finance long-term investments. However, researchers should empirically validate this assumption prior to automatic acceptance. Using an extensive data set of about 16 million firm-year observations for European firms over the years 1998–2017, we have shown that firms use trade credits to fund long-term investments. Small firms are more likely than large firms to purchase fixed assets with trade credits and they purchase a higher percentage of intangible assets relative to tangible assets with trade credits.

Our short-run model suggests that Small firms use trade credit to purchase about 6% of tangible fixed assets. The range for Micro, Medium, and Large firms ranges from 12% to 15%. Regarding intangible fixed assets, Micro and Small fund about 14% of these purchases in the short-run using trade credit. The comparable values are 23% for Large firms and 30% for Medium firms. Thus, European firms of all sizes often find it convenient to purchase some fixed assets using trade credit financing.

In the long-run, trade credits fund about 1% of the tangible and intangible asset purchases of Large firms. Since Large firms generally have access to other sources of credit, they use trade credit as a temporary funding source and over time switch to less expensive sources of funds. In contrast, smaller firms without the same access to bank credit continue to rely upon trade credit to finance their long-term investments. In the long-run, Micro, Small, and Medium-sized firms finance from 4% to 6% of tangible fixed asset purchases with trade credits. For intangible fixed asset purchases, the percentages funded by trade credit are 5% for Medium firms, 9% for Small firms, and 10% for Micro firms. For smaller firms, trade credit is an important source of long-term funding and trade credit should enter the capital structure decision. For larger firms, the standard practice of not considering trade credit as a part of the long-term capital structure may be reasonable.

Regarding substitutability or complementarity between trade credits and debt, our analysis has produced mixed results. In the short-run, trade credits and short-term bank loans are complementary sources of financing for Small, Medium, and Large firms. For Micro firms, the relationship is not significant, perhaps because of the difficulty in obtaining bank credit. Trade credit and long-term debt are short-run substitutes for Micro and Medium firms, but the relationship is not significant for other firms. Our long-run target adjustment analysis suggests that trade credits and short-term bank loans are complements for Medium-sized firms. In contrast, trade credits and long-term debt are substitutes for Micro firms, while the relationship is insignificant for other firms. These results suggest that the financing behavior of Micro firms is different from other firms.

The study's findings have practical implications for financial managers, particularly those in smaller firms. Managers should recognize trade credit as a viable and important long-term financing source. Understanding the optimal utilization of trade credits and their interaction with other financing sources can help optimize the cost of capital and enhance financial performance. Future research on trade credit financing could determine whether our conclusions hold for other regions of the world. Additional research is also needed to examine industry-specific differences in trade credit usage and to further explore the substitutability and complementarity relationships between trade credits and other sources of funds.

Funding statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability statement

The authors do not have permission to share data.

CRediT authorship contribution statement

Jan Bartholdy: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software. Dennis Olson: Formal analysis, Methodology, Validation, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.