Journal of Banking & Finance 33 (2009) 300–307

Contents lists available at ScienceDirect

Journal of Banking & Finance

journal homepage: www.elsevier.com/locate/jbf

Corporate trade credit and inventories: New evidence of a trade-off

from accounts payable and receivable
Spiros Bougheas a,*, Simona Mateut b, Paul Mizen a
a
School of Economics, University of Nottingham, University Park, Nottingham, UK
b
Department of Economics, University of Sheffield, England, UK

a r t i c l e i n f o a b s t r a c t

Article history: Trade credit is an important source of finance for firms and has been well researched, but the focus has
Received 16 April 2008 been on financial trade-offs. In this paper, we consider the trade-offs with inventories and develop a sim-
Accepted 17 July 2008 ple model that recognizes the incentives a firm faces to offer and receive trade credit. Our model identi-
Available online 16 September 2008
fies the response of accounts payable and accounts receivable to changes in the cost of inventories,
profitability, risk and liquidity, and importantly, this influence operates through a production channel.
JEL classification: Our results support the model and complement many existing studies focused on explaining the financial
G31
terms of trade credit.
G32
Ó 2008 Elsevier B.V. All rights reserved.
Keywords:
Trade credit
Inventories

1. Introduction 1987; Biais and Gollier, 1997), discrimination arguments (Brennan

et al., 1988), monitoring advantages (Jain, 2001; Mateut et al.,
Trade credit is a major element of corporate finance. Rajan and 2006), insurance (Cunat, 2007), product quality (Lee and Stove,
Zingales (1995) document that the volume of trade credit in aggre- 1993; Long et al., 1993), bankruptcy (Frank and Maksimovic,
gate was a significant part (17.8%) of total assets for all American 2005; Wilner, 2000), opportunistic behavior (Burkart and Elling-
firms in the early 1990s. In Germany, France and Italy, trade credit sen, 2004) and externalities (Daripa and Nilsen, 2005). Empirical
represents more than a quarter of total corporate assets, while in studies explore the relationships between accounts payable and
the United Kingdom 70% of total short-term debt (credit extended) accounts receivable and other balance sheet variables to corrobo-
and 55% of total credit received by firms is made up of trade credit rate or refute these theories and examine in detail the terms and
(Kohler et al., 2000; Guariglia and Mateut, 2006). Trade credit is conditions of trade credit.1 These are mostly financial theories of
also important in emerging economies, like China, where firms trade credit.
get limited support from the banking system (Ge and Qiu, 2007). Economists have become accustomed to considering separately
Accordingly, trade credit has been thoroughly researched but the transactions that involve the exchange of goods from those that in-
research has focused mainly on the financial substitutes and com- volve financial transactions. The separation is motivated by the
plements to trade credit by asking what is the advantage of obtain- benefits obtained from skill specialization: financially constrained
ing credit directly from sellers compared to other (often cheaper) buyers obtain funds in financial markets which they then use to
forms of credit such as bank loans. Atanasova and Wilson (2004) buy goods from sellers in goods markets. But the trade credit
find that to avoid bank credit rationing, smaller UK companies in- bridges goods and financial markets, and there is more to that
crease their reliance on inter-firm credit and Guariglia and Mateut bridge than the comparison of the relative costs of alternative
(2006) suggest the uptake of trade credit weakens the credit chan- forms of finance to firms or the terms of trade credit agreements.
nel as firms substitute trade finance for bank loans when monetary We argue that fresh motivation can be found for the study of trade
policy tightens. credit based on the advantage to the seller in inventory manage-
The literature provides many possible explanations for uptake ment from offering trade credit to the buyer. The advantage is that
or offer of trade credit based on informational asymmetries (Smith,
1
See for example, Mian and Smith (1992), Rajan and Zingales (1995), Petersen and
* Corresponding author. Tel.: +44 115 8466108; fax: +44 115 9514159. Rajan (1997), Ng et al. (1999), Demirguc-Kunt and Maksimovic (2002), Alphonse et al.
E-mail addresses: [email protected] (S. Bougheas), S.Mateut@ (2003), Fisman and Love (2003), Giannetti (2003), Preve (2003), Burkart et al. (2005),
sheffield.ac.uk (S. Mateut), [email protected] (P. Mizen). Cunningham (2004) and Love et al. (2005).

0378-4266/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankfin.2008.07.019
 S. Bougheas et al. / Journal of Banking & Finance 33 (2009) 300–307 301

the seller, who faces an uncertain demand for the product, may ex- ity and risk, and allowing for scale of bank borrowing. The effect of
tend trade credit to financially constrained customers in order to inventories on trade credit is dependent on firm size, as inventory
obtain credit-financed sales rather than accumulate costly invento- holding costs fall with size. These results have not been published
ries of finished goods, which may or may not be sold for cash in the before and provide evidence in favour of an inventory management
next period. This trade-off that the firm faces between inventories motive for offering trade credit.
and trade credit is the focus of this paper. In the following section, we develop a simple model that cap-
Our approach provides a theoretical two-period stochastic de- tures the trade-off between trade credit and inventory under sto-
mand model of firm behavior. Firms produce goods for sale, hold chastic demand. In Section 3, we present our empirical work and
inventories of goods that were produced but unsold at a cost, in the final section we conclude.
and, critically, offer and receive trade credit in the middle of a cred-
it chain. Therefore, producers facing an uncertain demand for their 2. Inventories and trade credit
products face an incentive to extend trade credit to their financially
constrained customers in order to promote sales rather than accu- Consider the following 2-period snapshot in the life of a single-
mulate costly inventories of finished goods. This incentive is lim- product firm that belongs in a competitive industry and lies in the
ited only by the need to obtain liquidity to meet their own middle of a trade credit chain. In period 1, when the firm decides
obligations, producers might readily offer trade credit on appropri- its level of production it faces uncertainty about the price for its
ate terms to enhance cash sales and boost demand. This trade-off product. The uncertainty can be related to both firm-specific
has not previously been explored in the economics literature.2 shocks and market instability. Let A denote the state of the world
We view the analysis of the trade-off between inventories and in period 2 and p(A) the corresponding price, where p0 (A). Further-
trade credit as complementary to theories that address financial ^ denote its level of production in period 1 and q (6 q
more, let q ^) de-
aspects of trade credit. Some early work on trade credit following note sales in period 2.4 Given that potential buyers are financially
a transactions costs approach has analyzed the trade-offs between constrained in period 2 the firm faces the following trade-off. It
the costs of financial transactions and the costs related to the ex- can avoid holding costly inventories by extending trade credit to
change of goods (see, for example, Nadiri (1969), Schwartz its customers, but trade credit is itself costly as the firm foregoes
(1974), Ferris (1981) and Emery (1987)). Only Emery (1987) con- cash with which to repay its own creditors. By offering goods on
siders explicitly the trade-off between trade credit and inventories credit the firm is trading-off potential future cash sales opportuni-
but does so within a deterministic variable demand framework. ties. We also assume that on average inventories are sold on dis-
More recently, Daripa and Nilsen (2005) have theoretically exam- count. The following maximization program captures this trade-off
ined how this trade-off influences the terms of trade credit agree- and solves for the optimal level of sales in period 2:
ments. In their model suppliers offer trade credit as an incentive to
buyers to hold higher inventories – shifting inventories from seller MaxfPðAÞq þ pE ðq
^  qÞ  vðq
^  q; xÞ  rðpðAÞq  cR ðmÞÞg:
q
to buyer. The underlying rationale for trade credit has some simi-
larities with ours when we consider a firm that lies in the middle of The first term represents sales in period 2 (both on cash and on
a credit chain, since suppliers reduce inventories by offering trade trade credit) while the second term captures future revenues from
credit and firms that accept trade credit from their suppliers and the sale of inventories, where pE denotes the expected future price.
v(,) represents the holding cost of inventories, q ^  q, and x is a shift
thus increase their inventories are also in the position to offer trade
credit to their own customers. In fact, the predictions of their mod- parameter that captures other factors that influence the cost of
el with respect to the effects of changes in inventories and profit inventories. The final term r() captures the cost of extending trade
margins on the levels of trade credit are the same as ours. credit (accounts receivable) that depends on the amount of trade
Our stylized model that provides directly testable predictions credit extended, which in turn is equal to sales minus cash receipts
on the response of accounts payable and accounts receivable to (assumed here to be directly related to the level of liquidity, m).5
changes in the cost of inventories, profitability, risk and liquidity, We impose the following restrictions on these functions: v1 > 0,
0

which operate by influencing production. Even the influence of v11 > 0, v2 > 0, v12 > 0, 1 > r > 0, r00 > 0 and c0R > 0. Thus, we assume
bank loans on trade credit operates by allowing greater production, that inventory costs are convex in the level of inventories and that
inventories and sales, financed in part through credit. Our contri- the shift parameter represents a firm characteristic that is associated
bution empirically is to directly test the predictions of our model with higher inventory costs. We further assume that costs related to
using GMM estimates in first-differences on an unbalanced panel extending trade credit (cost of receivables) are increasing at an
of UK firms drawn from FAME that includes larger FTSE-quoted increasing rate with the level of trade credit reflecting costs related
firms and those on the smaller AIM/OFEX exchange, as well as un- to lack of cash (higher demand for expensive accounts payable) and
quoted firms.3 The results show a direct influence of inventories on higher expected bankruptcy costs. Finally, firms that target a higher
accounts receivable but a negligible effect on accounts payable even liquidity will be less willing to offer trade credit.
after controlling for firms characteristics such as liquidity, profitabil- The f.o.c. of the above program is

pðAÞ  pE þ v1  pðAÞr 0 ¼ 0; ð1Þ

2
Two well cited theories of trade credit by Biais and Gollier (1997) on signalling that implicitly provides a solution for desired sales as a function of
and Burkart and Ellingsen (2004) on diversion, we find that both focus on the
the state of the world, q(A). Actual sales are restricted by produc-
relationship between trade credit and bank loans but neither one explicitly analyzes
the role of inventories. Operations research by contrast has recognized the link tion, thus optimal sales, q*, are given by
between trade credit and inventories but with an interest in inventory management
per se. While the inventory literature acknowledges the interaction between trade
^
qðAÞ 6 q q ¼ qðAÞ
If then ð2Þ
credit and inventories for optimal control, it is less interested in the economic ^
qðAÞ > q q ¼ q
^:
question of how a firm might consider trade-offs between greater overall sales and
lower inventory costs versus lower liquidity. In this paper, we take a closer look at
these trade-offs by proposing a theory and offering some evidence. See Robb and
4
Silver (2006) who provide an extensive review of the literature that explores the In a multi-period model sales would be restricted by the sum of production and
advantages of alternative inventory control methods subject to the availability of past inventories. For our purposes, setting past inventories equal to zero is
trade credit. inconsequential.
3 5
The source is the FAME (Bureau van Dijk) database collected by Electronic Accounts receivable, defined as sales minus cash receipts is directly observable in
Publishing. See http://fame.bvdep.com. the data.
302 S. Bougheas et al. / Journal of Banking & Finance 33 (2009) 300–307

The implicit function theorem implies that b  cp ðmÞ;

P  zqð AÞ ð5Þ
0 0 00
dq p ðAÞð1  r  r pðAÞqÞ and (expected) accounts receivable
¼ ; ð3Þ
dA v11 þ ðpðAÞÞ2 r00 Z Z
R pðAÞqðAÞf ðAÞdA þ b ðAÞdA  cR ðmÞ:
pðAÞqð AÞf ð6Þ
where the second-order condition for a maximum implies that the A<b
A A>b
A
denominator is positive. Whether sales increase with the state of b will affect both ac-
One useful observation is that a change in A
demand would depend on the cost of extending trade credit. As long
counts payable and accounts receivable in the same direction. As
as the corresponding function is not too convex (r11 is low) an in-
production increases, for a given level of liquidity, both trade credit
crease in the state of demand would imply higher sales.6 From
dq terms go up.
now on we assume the more plausible case dA > 0. Then, together,
(2) and (3) imply that there exists a state of the world A b such that
b In low demand states, A < A,b the firm sells less than its out- 2.1. Cost of holding inventories
^ ¼ qð AÞ.
q
put and thus inventories increase while in high demand states,
b the firm offers sufficient trade credit so that its entire output Inventory costs, such as warehousing and stockout costs, are
A > A,
not directly observable but the parameter x in the inventory cost
is sold.
function captures firm characteristics that might be related to
Next, we solve for the optimal level of output in period 1. Let z
these costs. Shirley and Winston (2004), in their econometric spec-
denote the constant marginal cost. The firm uses its liquidity, m, to
ification of the inventory cost function, in addition to the level of
pay part of its cost of production, zq^, and to cover the rest it bor-
inventories have included industry and location dummies to cap-
rows from its suppliers.7,8 In period 1 the firm solves the following
ture variations in these costs due to variations in commodity type
program:
Z and geographic location. As Fazel (1997) has argued the size of the
b  qðAÞÞ  vðqð AÞ
b  qðAÞ; xÞ firm might also be important as smaller firms have less flexibility
Max ½pðAÞqðAÞ þ pE ðqð AÞ
bA A<b
A in their choice of purchasing methods.9
Z From (4), we get
 rðpðAÞqðAÞ  cR ðmÞÞf ðAÞdA þ b  rðpðAÞqð AÞ
½pðAÞqð AÞ b R
A>b b
dA v12 f ðAÞdA
A<b
A
A
b  bðzqð AÞ
b  cp ðmÞÞ; ¼ < 0:
 cR ðmÞÞf ðAÞdA  zqð AÞ dx D

where the last term, b(), captures the cost of holding accounts pay- Thus firms that face higher inventory costs at the margin will carry
able which are equal to the cost of production, zq ^, minus cash pay- less accounts receivable and less accounts payable.
ments cp(m) (where we assume that c0p > 0; i.e. firms with higher Because of its static framework, our model does not distinguish
liquidity avoid costly accounts payable) and f() is the density func- between stocks and flows. However, our model clearly suggests
00
tion of the distribution of A. We assume that b0 > 0 and b > 0 and that any stock of inventories carried forward will have the same
they capture costs that might be related to the deterioration of marginal effects as x. In this case we would write the inventory
function as vðINV 1 þ q^  q; xÞ where INV1 denotes lagged inven-
the balance sheet as trade credit expands.
The f.o.c. of the period-1 program is tories. Once more using (4) we get:
Z Z R
b
dA v11 f ðAÞdA
0 A<b
½pE  v1 f ðAÞdA þ ½pðAÞð1  r 0 Þf ðAÞdA  zð1 þ b Þ ¼ 0: ¼ A
< 0;
A<b
A A>b
A dINV 1 D
ð4Þ then our model predicts that firms with a higher stock of invento-
Higher production increases inventories in low demand states and ries will show lower levels of accounts receivable and accounts
consequently increases both inventory costs and future revenues payable.
(first term) and revenues in high demand states (second term).
The last term captures the effect of a marginal increase in output 2.2. Changes in profitability and risk
on production and financing costs. The above first-order condition
implicitly provides a solution for optimal output as a function of Even firms that operate in the same industry and are of similar
various exogenous variables. size can differ in terms of profitability and have different risk rat-
For our comparative statics below the total derivative of (4), D, ings. This could be because of variations in technologies, organiza-
b will be useful:
with respect to A, tion and past financing policies.10 One way to capture these
Z Z  differences in our model is by allowing for changes in the distribu-
D v11 f ðAÞdA þ
00
ðpðAÞÞ2 r 00 f ðAÞdA þ z2 b dq ^ < 0; tion of the state of the world. A change in the mean keeping the var-
A<b
A A>b
A iance the same (so that the distribution with the higher mean
where (1) was used for simplifying the above derivation. The com- dominates the other in the first-order-stochastic-dominance sense)
parative statics of our model that we are interested in are the effects represents changes in profitability, while a change in the variance
of changing a number of exogenous variables on the levels of ac- keeping the mean the same (mean-preserving spreads) captures
counts payable changes in riskiness. Once more, it is clear from (4) that the effect
of any change in the distribution on accounts payable and expected
accounts receivable will be through changes in output and that any
6
The reason that the sign is ambiguous is because an increase in the state of
change in the distribution will affect accounts payable and expected
demand (higher A which implies a higher p) will boost revenues even if sales stay the
same. But this would imply that the firm would have to offer more trade credit. If the accounts receivable in the same direction. Actually, we can show
cost of the latter is too high it might decide to lower sales.
7
Notice that m denotes both liquidity in period 2 that includes cash receipts and
9
available liquidity in period 1. This simplification is deemed necessary because of data The inventory control argument employed tells us small firms opt more often for
constraints. Trade credit is held on average for periods much shorter than the yearly the economic order quantity (EOQ) purchasing option, which requires higher
frequency of our data. Thus, m captures the average liquidity over a 12-month period. inventories, because they cannot effectively implement the just-in-time (JIT)
8
Of course, firms have other short-term financial options that for the moment we alternative.
10
ignore so that we can concentrate on the trade-off between trade credit and The presence of trade credit itself in its balance sheet can also affect a firm’s credit
inventories. risk rating; see for example Ebnöther and Vanini (2007) and Carling et al. (2007).
 S. Bougheas et al. / Journal of Banking & Finance 33 (2009) 300–307 303

that an increase in profitability will have a positive effect on both ac- represents firm’s gross liquid assets (cash, bank deposits, and other
counts and an increase in riskiness can have either a positive or a current assets excluding accounts receivables), and Banksit, which
negative effect on the two accounts. represents short-term bank loans.12 With the exception of Riskit all
Consider first a change in profitability on (4), keeping A b at its variables are scaled by total sales.
optimal value before the change. The change subtracts mass from We expect the use of trade credit to differ from industry to
the first integral and adds mass on the second integral having an industry since empirical studies have found wide variations across
overall positive effect on the left-hand side of (4). This implies that industries but rather similar credit terms within industries (Burk-
the optimal value of A, b and thus output, must be higher, which in art et al., 2005; Ng et al., 1999; Nilsen, 2002). At the same time, the
turn implies that both accounts payable and accounts receivable reliance of firms on internal finance relative to external finance fol-
will move up. lows an industry pattern. In addition, as Shirley and Winston
Next, consider an increase in riskiness. Again keeping A b at its (2004) suggest, inventory costs differ significantly across indus-
optimal value before the change, the increase in dispersion will have tries. This is why we allocate firms to one of the following nine
the following effects on (4). It will subtract mass from both integrals manufacturing industrial sectors: metals and metal goods; other
that implies a negative effect on the left-hand side of (4) since the minerals, and mineral products; chemicals and man made fibres;
value of the two integrals together is positive. It will add mass on mechanical engineering; electrical and instrument engineering;
the left tail of the first integral and at the margin this effect will de- motor vehicles and parts, other transport equipment; food, drink,
crease the left-hand side of (4) while it will also add mass on the and tobacco; textiles, clothing, leather, and footwear; and others
right tail of the second integral which at the margin will increase (Blundell et al., 1992). In our specifications, we control for the
the left-hand side of (4). Without any further knowledge of the dis- industry characteristics by including industry dummies interacted
tribution function we cannot determine the sign of the overall effect with time dummies. Thus the inventory costs can differ between
although it seems that both effects will more likely be negative. industries and across time.
In order to check whether the sensitivity of trade credit usage
2.3. Changes in liquidity (both extended and received) differs at firms with different size,
which also affects the costs of holding inventories (Fazel, 1997),
We have assumed that a firm’s level of liquidity affects the cost we define the variable Sizeit as the logarithm of firm’s real assets.
of both payables and receivables. Here, we consider how changes We then interact it with the Stocksit variable to control for cost dif-
in liquidity affects the levels of output and trade credit. Once more ferences in holding inventories. We postulate that holding costs
using (4) we have decrease with the level of inventories but also with the size of
R 00 the firm. Therefore, the estimated equations take the following
b 
dA pðAÞr00 c0R ðmÞf ðAÞdA  zb c0p ðmÞ
A>b
A form:
¼ > 0:
dm D
ARit Stocksit Stocksit
So an increase in liquidity has a positive effect on production. Then ¼ ai þ b þ  Sizeit b2 þ Riskit b3
Salesit Salesit 1 Salesit
the effect of a change in liquidity on payables is given by
Profitsit Liquidit Banksit
b þ b þ b þ b þ Sizeit b7 þ eit ;
dP ^ dA
dq Salesit 4 Salesit 5 Salesit 6
¼z  c0p ðmÞ;
dm b
dA dm and
and on receivables by APit Stocksit Stocksit Profitsit
Z ¼ ai þ c þ  Sizeit c2 þ Riskit c3 þ c
dE½R dq b
^ dA Salesit Salesit 1 Salesit Salesit 4
¼ pðAÞf ðAÞdA  c0R ðmÞ: Liquidit Banksit
dm A>b
A b dm
dA þ c þ c þ Sizeit c7 þ uit ;
Salesit 5 Salesit 6
These results capture both the indirect effects of higher liquidity on
the two accounts through its influence on their respective costs and where ai is a firm-specific component, bi’s and ci’s are coefficient
the direct effects on cash receipts and cash payments. If the direct values, and eit and uit are the respective idiosyncratic error compo-
effects dominate then higher liquidity will have a negative impact nent. We control for firm-specific, time-invariant effects, and for the
on both accounts. possible endogeneity of the regressors, by using a first-difference
GMM approach.13 Lags of each of the regressors (including the inter-
3. Empirical methodology and data characteristics action terms) are used as instruments.14 Both time dummies and
industry dummies interacted with time dummies are included in
To test the predictions of our model we define our dependent all our regressions. We report both the first- (m1) and the second-or-
variables AR and AP to represent accounts receivable (defined as der (m2) test for serial correlation, which are asymptotically distrib-
the balance sheet item trade debtors) and accounts payable (de- uted as a standard normal under the null of no serial correlation of
fined as the balance sheet item trade creditors). We explain both the differenced residuals. At the same time, the variables in the
trade credit extended and trade credit received with the same instrument set should be uncorrelated with the error term in the rel-
independent variables: Stocksit, the level of finished goods and evant equation if the model is correctly specified. We report the Sar-
work in process inventories; Riskit, which measures the likelihood gan (Hansen) test for the legitimacy of variables dated t-2 as
of company failure in the twelve months following the date of instruments in the differenced equation. Under the null of instru-
calculation, where a lower value indicates that the firm is more ment validity the Sargan test for overidentifying restrictions is
risky.11 Profitsit gives the firm’s profit (or loss) for the period, Liquidit
12
We include the latter variable in our specifications as we think that firms’ use of
trade credit relies heavily on their use of bank loans even though our theoretical
11
We are using the quiscore indicator produced by Qui Credit Assessment Ltd., model has concentrated on trade credit only. We extend the theoretical model to
which measures the likelihood of company failure in the twelve months following the include bank loans and explain our empirical results in the appendix.
13
date of calculation. Quiscore is given as a number in the range from 0 to 100. The All our regressions are performed in Stata using the command xtabond2
lower its quiscore the more risky a firm is likely to be. This is a wider definition of developed by Roodman (2005).
14
perceived financial health than the commonly used bond rating, which only applies to This is the reason why the size of the sample in the tables of results is smaller
rated firms (Whited, 1992; Kashyap and Stein, 1994). than the full sample.
304 S. Bougheas et al. / Journal of Banking & Finance 33 (2009) 300–307

asymptotically distributed as a chi-square with degrees of freedom Table 2

equal to the number of instruments less the number of parameters. Accounts receivables

Our dataset is derived from the profit and loss and balance sheet 1 2 3 4
data gathered by Bureau Van Dijk Electronic Publishing in the Stocks 0.369 ***
0.554 ***
1.066 **
1.444***
FAME database, which provides information on companies over a (0.119) (0.146) (0.415) (0.547)
ten year period. Our sample includes firms operating in the manu- Stocks*size 0.085** 0.109*
facturing sector and covers the period 1993–2003. The majority (0.040) (0.061)
Risk 0.138*** 0.210*** 0.148*** 0.203***
(over 99%) of the firms included in the dataset are not traded on
(0.029) (0.037) (0.030) (0.036)
the stock market. The large proportion of unquoted firms means Profits 0.268*** 0.493*** 0.284*** 0.488***
we are likely to observe many firms that are financially constrained (0.104) (0.099) (0.100) (0.094)
i.e. they are unable to obtain resources from capital markets and Liquid 0.091** 0.071 0.120*** 0.082*
(0.043) (0.051) (0.039) (0.044)
possibly also banks to purchase inputs for production. Therefore
Banks 0.099*** 0.097** 0.083*** 0.091**
the relevant consideration for these firms is the opportunity to (0.033) (0.041) (0.031) (0.037)
purchase goods at the margin financed by additional trade credit Size 0.030*** 0.019
or bank loans. We excluded companies that changed the date of (0.009) (0.013)
their accounting year-end by more than a few weeks so that the
Observations 56,432 56,432 56,432 56,432
data refer to 12 month accounting periods. We excluded observa-
No. of firms 10,877 10,877 10,877 10,877
tions in the 1% tails for each of the regression variables to control m1(p) 0.00 0.00 0.00 0.00
for the potential influence of outliers. Table 1 reports summary m2(p) 0.19 0.69 0.18 0.71
statistics. Hansen/Sargan (p) 0.24 0.39 0.33 0.44

Notes: All specifications are estimated using a GMM first-difference specification.

3.1. Empirical results Test statistics and standard errors (in parentheses) are asymptotically robust to
heteroskedasticity. m1 (m2) is a test for first- (second-) order serial correlation in
Tables 2 and 3 report the relationship between accounts receiv- the first-differenced residuals, asymptotically distributed as N(0,1) under the null of
no serial correlation. The Hansen/Sargan test is a test of the overidentifying
able (AR) and accounts payable (AP) for firms in our panel and their restrictions, distributed as chi-square under the null of instrument validity.
characteristics. Column 1 in both tables presents regression results Instruments include Stocksi,t-2; Stocksi,t-2*Sizei,t-2; Riski,t-2; Profiti,t-2; Liquidityi,t-2;
ignoring the influence ofSizeit while the remaining columns in- Banksi,t-2; and further lags. Time dummies and time dummies interacted with
clude Sizeit as an additional variable (column 2), as an interaction industry dummies were always included as regressors and as instruments. *, **, ***
indicate significance at the 10%, 5%, and 1% level, respectively.
term with Stocksit (column 3) and both as an additional variable
and an interaction term (column 4).
The level of inventories is predicted to have a direct negative af- As proposed in the model, there are interactions that influence
fect on AR, and an indirect negative affect on AP from our model. the scale of the impact of inventory holding costs. The most impor-
Here production decisions are critical, since when production ex- tant of these is the size of the firm, which has an impact on both AR
ceeds sales causing inventories to increase, other things equal, and AP. Larger firms both extend and receive more trade credit to
firms will have an incentive to offer more trade credit in order to and from their business partners (even after scaling by sales) and
gain more total sales and hold fewer inventories. The argument they also have lower inventory holding costs within the same
is similar to the sales motive identified by Wilson and Summers industry other things equal. The levels of AR and AP divided by
(2002), where firms extend sales by offering goods on account in sales increase with the size of the firm since the additional size var-
the first instance. Our results show that inventories have a large, iable is significant and positive in both regressions (column 2, Ta-
negative and significant impact on AR and a negligible impact on bles 2 and 3) and when we interact the Stocks variable with the
AP. measure for size (columns 3 and 4, Tables 2 and 3) it becomes

Table 1
Summary statistics (means and standard deviations)

Variable Whole sample Small bottom 75% Large top 25% Small bottom 50% Large top 50% Small bottom 25% Large top 75%
TD 0.171 0.176 0.157 0.176 0.166 0.174 0.171
(0.078) (0.077) (0.077) (0.078) (0.077) (0.078) (0.077)
TC 0.107 0.108 0.101 0.109 0.104 0.109 0.106
(0.062) (0.062) (0.059) (0.063) (0.061) (0.063) (0.061)
Stocks 0.121 0.118 0.131 0.112 0.130 0.102 0.128
(0.089) (0.089) (0.088) (0.087) (0.090) (0.082) (0.090)
Risk 0.540 0.535 0.557 0.525 0.556 0.507 0.551
(0.214) (0.212) (0.220) (0.209) (0.218) (0.207) (0.215)
Profits 0.031 0.029 0.036 0.028 0.034 0.028 0.032
(0.075) (0.071) (0.087) (0.070) (0.080) (0.070) (0.077)
Liquid 0.155 0.131 0.225 0.116 0.194 0.100 0.173
(0.203) (0.177) (0.254) (0.160) (0.231) (0.145) (0.216)
Banks 0.152 0.135 0.201 0.125 0.179 0.115 0.164
(0.195) (0.168) (0.253) (0.151) (0.228) (0.138) (0.209)
Size 8.695 8.044 10.654 70.587 90.805 70.074 90.237
(1.483) (0.892) (1.144) (0.688) (10.210) (0.559) (10.287)
Observations 72,905 54,714 18,191 36,477 36,428 18,267 54,638

Note: TD represents trade debit (accounts receivables) and TC represents trade credit (accounts payables). Stocks stands for stocks of inventories; Risk measures the likelihood
of company failure in the twelve months following the date of calculation, where a lower value indicates that the firm is more risky. Profits gives the firm’s profit (or loss) for
the period; Liquid represents firm’s liquid assets (cash, bank deposits, and other current assets), Banks represents short-term bank loans. Size is the logarithm of total real
assets. With the exception of Risk and Size all variables are scaled by total sales. We separate firms into two size categories using a dummy variable for size named Large,
which takes value 1 in a given year if the firm’s total assets are in the top 25 (columns 2–3), top 50 (columns 4–5), and top 75 (columns 6–7) percentile of the distribution of
total assets of all the firms in that particular industry and year. Thus firms are allowed to transit between categories.
 S. Bougheas et al. / Journal of Banking & Finance 33 (2009) 300–307 305

Table 3 tween avoiding holding costs of inventories and obtaining future

Accounts payables cash sales, the cost of holding stocks is lower for larger firms. As
1 2 3 4 before, AP are not influenced by the firm’s stocks of inventories
Stocks 0.038 0.109 0.035 and our other results do not change when we drop this variable
(0.079) (0.092) (0.273) from the regression (column 4, Table 3).
Stocks*Size 0.004 Risk, profitability and liquidity have an indirect influence on AR
(0.026) and AP through the level of production and inventories. Our model
Risk 0.167*** 0.200*** 0.157*** 0.224***
(0.028) (0.035) (0.026) (0.036)
predicts profitability will increase both AR and AP, but the signs of
Profits 0.231*** 0.344*** 0.180*** 0.361*** risk and liquidity are not determined. However, as we remarked
(0.078) (0.076) (0.069) (0.078) earlier we have reasons to think that our Risk variable might have
Liquid 0.072*** 0.089*** 0.045* 0.079** a negative impact on both AR and AP and this is what our empirical
(0.027) (0.031) (0.023) (0.032)
results suggest. We have found a negative relationship between
Banks 0.109*** 0.116*** 0.104*** 0.167***
(0.033) (0.040) (0.030) (0.045) liquidity and trade credit extended, and this is consistent with Pet-
Size 0.016** 0.018*** ersen and Rajan (1997), who also found a negative relationship be-
(0.006) (0.007) tween AR and liquidity. Profitability has a positive effect since
Observations 55,848 55,848 55,848 55,848 extra profit can be channeled towards AR, and trade credit is more
No of firms 10,806 10,806 10,806 10,806 likely to be offered to profitable firms. The common sign pattern is
m1(p) 0.00 0.00 0.00 0.00 supported in our results and profitability is positive in both AR and
m2(p) 0.44 0.19 0.62 0.14
AP regressions. These firm-specific characteristics were also found
Hansen/Sargan(p) 0.52 0.60 0.15 0.86
to be important in determining access to bank loans in Bougheas
Notes: All specifications are estimated using a GMM first-difference specification. et al. (2006) but our contribution here is to show that besides indi-
Test statistics and standard errors (in parentheses) are asymptotically robust to
rectly influencing AP by relaxing credit limits from banks, they
heteroskedasticity. m1 (m2) is a test for first- (second-) order serial correlation in
the first-differenced residuals, asymptotically distributed as N(0,1) under the null of have direct effects on AP and AR through inventories since the firm
no serial correlation. The Hansen/Sargan test is a test of the overidentifying faces a trade-off between incurring holding costs of inventories
restrictions, distributed as chi-square under the null of instrument validity. with the possibility of future cash sales versus sales on credit now.
Instruments include Stocksi,t-2; Stocksi,t-2*Sizei,t-2; Riski,t-2; Profiti,t-2; Liquidityi,t-2;
We introduce the variable Banksit as a control variable to deter-
Banksi,t-2; and further lags. Time dummies and time dummies interacted with
industry dummies were always included as regressors and as instruments. *, **, ***
mine the effect of bank loans on AR and AP. We find that it in-
indicate significance at the 10%, 5%, and 1% level, respectively. creases AR and reduces AP, which is consistent with the pecking
order of finance view, which assumes that trade credit is more
expensive than bank loans. The idea that trade credit is lower
apparent that as the size of the firm increases, stocks of inventories down the pecking order of finance has been supported by Petersen
play a lesser role in the firm’s decision to extend AR since the po- and Rajan (1997) who find evidence that US firms increase AP
sitive coefficient of the interacted term offsets the negative effect when credit is rationed i.e. there is restricted access to bank loans
of stocks on AR (column 3, Table 2). In terms of the trade-off be- and capital markets. It is also consistent with the assumptions of

Table 4
Robustness results. Results for accounts receivables (AR) are presented in columns 1–2 and for accounts payables (AP) in columns 3–5

AR AP
1 2 3 4 5
Stocks 0.375*** 0.041
(0.119) (0.080)
Stocks*small 0.354*** 0.012
(0.124) (0.083)
Stocks*large 0.226** 0.064
(0.103) (0.066)
Risk 0.138*** 0.134*** 0.166*** 0.184*** 0.159***
(0.029) (0.028) (0.028) (0.029) (0.027)
Profits 0.273*** 0.241** 0.234*** 0.251*** 0.177***
(0.103) (0.096) (0.078) (0.083) (0.069)
Liquid 0.088** 0.105*** 0.072*** 0.071** 0.056**
(0.043) (0.039) (0.027) (0.028) (0.024)
Banks 0.098*** 0.089*** 0.109*** 0.141*** 0.104***
(0.033) (0.031) (0.033) (0.036) (0.031)
Large 0.013*** 0.006** 0.006**
(0.003) (0.002) (0.003)
Observations 56,432 56,432 55,848 55,848 55,848
No. of firms 10,877 10,877 10,806 10,806 10,806
m1(p) 0.00 0.00 0.00 0.00 0.00
m2(p) 0.22 0.16 0.41 0.38 0.69
Hansen/Sargan(p) 0.25 0.37 0.54 0.82 0.19

Instead of using the continuous variable Size, we consider the situation of each firm relative to that of other firms in the industry in which that firm operates and for each year.
We define a dummy variable for size named Large, which takes value 1 in a given year if the firm’s total assets are in the top 25 percentile of the distribution of total assets of
all the firms in that particular industry and year. This way we allow firms to transit between categories. We then interact Stocks with (1-Large) and Large to capture the
impact of costs of holding inventories separately for small and large firms.
Notes: All specifications are estimated using a GMM first-difference specification. Test statistics and standard errors (in parentheses) are asymptotically robust to heter-
oskedasticity. m1 (m2) is a test for first- (second-) order serial correlation in the first-differenced residuals, asymptotically distributed as N(0, 1) under the null of no serial
correlation. The Hansen/Sargan test is a test of the overidentifying restrictions, distributed as chi-square under the null of instrument validity. Instruments include Stocksi,t-2;
Stocksi,t-2*Smalli,t-2; Stocksi,t-2*Largei,t-2; Riski,t-2; Profiti,t-2; Liquidityi,t-2; Banksi,t-2; and further lags. Time dummies and time dummies interacted with industry dummies were
always included as regressors and as instruments. *, **, *** indicate significance at the 10%, 5%, and 1% level, respectively.
306 S. Bougheas et al. / Journal of Banking & Finance 33 (2009) 300–307

Burkart et al. (2005). We offer a theoretical extension to our model our model using GMM estimates in first-differences on an unbal-
that explains how our findings can be embedded in our model in anced panel of UK firms drawn from FAME that includes larger
the Appendix, although it is a secondary consideration in relation FTSE-quoted firms and those on the smaller AIM/OFEX exchange,
to the role of inventories. as well as unquoted firms. Our results support the model suggest-
To summarize, our theoretical model explains the decision pro- ing that there is an inventory channel of trade credit, complement
cess of the firm, which derives optimal production and sales sub- many existing studies focusing on the financial terms of trade
ject to the state of the world. This indicates how AR and AP credit.
respond to changes in output as the state of the world improves,
and risk, profitability and liquidity as well as bank loans impinge Acknowledgement
on this state of the world, affecting production and inventories,
and therefore alter the levels of AP and AR. This is the inventory We are grateful to the Editor (Ike Mathur) for helpful comments
channel of trade credit which we confirm empirically. that improved the exposition of the paper.
Our final analysis involves exploration of the robustness of our
results. We consider firms that are relatively large or small in rela- Appendix. Introducing bank loans
tion to others in their industry by defining a dummy variable Large
which takes value 1 in a given year if the firm’s total assets are in In order to concentrate on the trade-off between trade credit
the top 25 percentile of the distribution of the total assets of all the and inventories we ignore any other forms of finance available to
firms in that particular industry and year. By contrast small firms firms when we developed our theoretical model, and we assume
are in the remainder of the distribution. We allow firms to transit with Petersen and Rajan (1997) and Burkart and Ellingsen (2004)
between categories and we also acknowledge that firm size is mea- that bank credit is cheaper than trade credit. We also assume the
sured specifically for each industry. The cut-off value is decided by firm faces a bank credit limit b
L, which may be related to credit con-
the distribution of the firms in our sample reported in Table 1b, straints arising from asymmetric information or any other cause,
where centile values for real assets are displayed. This underlines and we denote by L the amount actually borrowed. Then under
the fact that our dataset mainly comprises small firms. the supposition that bank loans are cheaper than trade credit for
In Table 4 we first investigate the effects of the Large dummy on a given maturity we have
the level of AR and AP. We find that large firms extend more AR
b  cp ðmÞ; Lg:
L ¼ minfzqð AÞ
and more AP, as indicated by a positive and significant coefficient
(columns 1, 3 and 4, Table 4). Then by interacting the Large and
Thus, the firm exhausts its bank credit limit before it seeks credit
Small = (1  Large) dummies with Stocksit we determine whether
from its suppliers.16 Accounts payable now are
inventories play a greater (lesser) role on AR and AP for large
(small) firms. We find that firms that are relatively large compared b  cp ðmÞ  L:
P  zqð AÞ
to other firms in the same industry and year respond less to inven-
For firms that are not financially constrained, L < L, increasing the
tories than do small firms. This clarifies the point that relatively
bank credit limit will not affect any of their decisions. In contrast,
large firms are less influenced by the trade-off between current
for financially constrained firms, L ¼ L, an increase in the bank cred-
credit sales and future cash sales because their holding costs are
it limit will have the following effects:
lower. These results are robust to the choice of the cut-off level
and are practically identical in terms of signs, and relative magni- b
dA zb
00

tudes, when setting the cut-off value at the 50th percentile, 70th ¼ > 0:
dL D
percentile, and 80th percentile.
Production will increase and thus accounts receivable will also in-
The results are qualitatively the same when we reconsider our
crease. In contrast, we have
analysis using data only for the unquoted firms, as they constitute
the majority of the firms in our sample. Results are also qualita- dP dq b
^ dA
tively the same when we eliminate from the analysis the larger ¼z  1:
dL b
d A dL
public firms. Finally, including only time dummies in the instru-
ment matrix and leaving out industry dummies interacted with So accounts receivable are complements to bank loans but accounts
time dummies produces similar results.15 payable can be either complements or substitutes.

4. Conclusions References

Alphonse, P., Ducret, J., Severin, E., 2003. When trade credit facilitates access to bank
We develop a simple theoretical model with a stochastic de- finance: Evidence from US small business data. Working Paper. University of
mand framework that captures the trade-off between inventories Lille.
Atanasova, C., Wilson, N., 2004. Disequilibrium in the UK corporate loan market.
and trade credit. The essential elements are that the firm is in
Journal of Banking and Finance 28, 595–614.
the middle of a credit chain, and produces goods for sale, holds Biais, B., Gollier, C., 1997. Trade credit and credit rationing. Review of Financial
inventories of goods that were produced but unsold at a cost and Studies 10, 903–937.
in the face of uncertain demand for its products extends trade Blundell, R., Bond, S., Devereux, M., Schiantarelli, F., 1992. Investment and Tobin’s Q:
Evidence from company panel data. Journal of Econometrics 31, 233–257.
credit to its financially constrained customers to obtain additional Bougheas, S., Mizen, P., Yalcin, C., 2006. Access to external finance: Theory and
sales. Our model provides directly testable predictions to identify evidence on the impact of monetary policy and firm-specific characteristics.
the response of accounts payable and accounts receivable to Journal of Banking and Finance 30, 199–227.
changes in the cost of inventories, profitability, risk and liquidity,
and importantly, this influence operates through a production 16
To keep things simple we have assumed that the limit is exogenous. Recent
channel. Even the influence of bank loans on trade credit operates developments suggest that the credit limit is increasing in tangible assets, T, and
by allowing greater production, inventories and ultimately sales fi- decreasing in existing debt, D, (Bougheas et al., 2006), increasing in past accounts
receivable as being one of the assets that banks are willing to accept as collateral (e.g.
nanced in part through credit. We directly test the predictions of Mian and Smith (1992), Burkart and Ellingsen (2004) and Voordeckers and Steijvers
(2007)) and past accounts payable following Biais and Gollier (1997) and Burkart and
Ellingsen (2004) who argue that the willingness of firms to supply trade credit
15
These results are not reported but are available on request. provides a signal to banks that the borrower is creditworthy.
 S. Bougheas et al. / Journal of Banking & Finance 33 (2009) 300–307 307

Brennan, M., Maksimovic, V., Zechner, J., 1988. Vendor financing. Journal of Kohler, M., Britton, E., Yates, T., 2000. Trade credit and the monetary transmission
Finance 43, 1127–1141. mechanism. Working Paper. Bank of England.
Burkart, M., Ellingsen, T., 2004. In-kind finance: A theory of trade credit. American Lee, Y., Stove, J., 1993. Product risk, asymmetric information, and trade credit.
Economic Review 94, 569–590. Journal of Financial and Quantitative Analysis 28, 285–300.
Burkart, M., Ellingsen, T., Giannetti, M., 2005. What you sell is what you Long, M., Malitz, I., Ravid, A., 1993. Trade credit, quality guarantees and product
lend? Explaining trade credit contracts. Finance Working Paper No 71/2005, marketability. Financial Management 22, 117–127.
ESGI. Love, I., Preve, L., Sarria-Allende, V., 2005. Trade credit and bank credit: Evidence
Carling, K., Jacobson, T., Lindé, J., Roszbach, K., 2007. Corporate credit risk modeling from recent financial crises. Policy Research Working Paper 3716. World Bank.
and the macroeconomy. Journal of Banking and Finance 31, 845–868. Mateut, S., Bougheas, S., Mizen, P., 2006. Trade credit, bank lending and monetary
Cunat, V., 2007. Trade credit: Suppliers as debt collectors and insurance providers. policy transmission. European Economic Review 50, 603–629.
Review of Financial Studies 20, 491–527. Mian, S., Smith, C., 1992. Accounts receivable management. Journal of Finance 47,
Cunningham, R., 2004. Trade credit and credit rationing in Canadian firms. Working 169–200.
Paper 2004-49, Bank of Canada. Nadiri, M., 1969. The determinants of trade credit in US total manufacturing sector.
Daripa, A., Nilsen, J., 2005. Subsidizing inventory: A theory of trade credit and Econometrica 37, 408–423.
repayment. Working Papers in Economics and Finance No. 0522. Birkbeck Ng, C., Smith, J., Smith, R., 1999. Evidence on the determinants of credit terms in
College. interfirm trade. Journal of Finance 54, 1109–1129.
Demirguc-Kunt, A., Maksimovic, V., 2002. Firms as financial intermediaries: Nilsen, J., 2002. Trade credit and the bank lending channel. Journal of Money, Credit
Evidence from trade credit data. Working Paper. University of Maryland. and Banking 34, 226–253.
Ebnöther, S., Vanini, P., 2007. Credit portfolios: What defines risk horizons and risk Petersen, M., Rajan, R., 1997. Trade credit: Theories and evidence. Review of
measurement? Journal of Banking and Finance 31, 3663–3679. Financial Studies 10, 661–697.
Emery, G., 1987. An optimal financial response to variable demand. Journal of Preve, L., 2003. Financial distress and trade credit: An empirical analysis. Working
Financial and Quantitative Analysis 22, 209–225. Paper. University of Texas.
Fazel, F., 1997. A comparative analysis of inventory costs of JIT and EOQ purchasing. Rajan, R., Zingales, L., 1995. What do we know about capital structure? Some
International Journal of Physical Distribution and Logistics Management 27, evidence from international data. Journal of Finance 50, 1421–1460.
496–504. Robb, D., Silver, E., 2006. Inventory management under date-terms supplier trade
Ferris, S., 1981. A transactions theory of trade credit use. Quarterly Journal of credit with stochastic demand and lead time. Journal of Operational Research
Economics 94, 243–270. 57, 692–702.
Fisman, R., Love, I., 2003. Trade credit, financial intermediary development and Roodman, D., 2005. Xtabond2: Stata module to extend xtabond dynamic panel data
industry growth. Journal of Finance 58, 353–374. estimator. Working Paper. Center for Global Development, Washington.
Frank, M., Maksimovic, V., 2005. Trade credit, collateral, and adverse selection. Schwartz, R., 1974. An economic model of trade credit. Journal of Financial and
Working Paper. University of Maryland. Quantitative Analysis 9, 643–657.
Ge, Y., Qiu, J., 2007. Financial development, bank discrimination and trade credit. Shirley, C., Winston, C., 2004. Firm inventory behavior and the returns from
Journal of Banking and Finance 31, 513–530. highway infrastructure investment. Journal of Public Economics 55, 398–415.
Giannetti, M., 2003. Do better institutions mitigate agency problems? Evidence Smith, J., 1987. Trade credit and informational asymmetry. Journal of Finance 42,
from corporate finance choices. Journal of Financial and Quantitative Analysis 863–876.
38, 185–212. Voordeckers, W., Steijvers, T., 2007. Business collateral and personal commitments
Guariglia, A., Mateut, S., 2006. Credit channel, trade credit channel, and inventory in SME lending. Journal of Banking and Finance 30, 3067–3086.
investment: Evidence from a panel of UK firms. Journal of Banking and Finance Whited, T., 1992. Debt, liquidity constraints and corporate investment: Evidence
30, 2835–2856. from panel data. Journal of Finance 47, 1425–1460.
Jain, N., 2001. Monitoring costs and trade credit. Quarterly Review of Economics and Wilner, B., 2000. The exploitation of relationships in financial distress: The case of
Finance 41, 89–110. trade credit. Journal of Finance 55, 153–178.
Kashyap, A., Stein, J., 1994. Monetary policy and bank lending. In: Mankiw, G. (Ed.), Wilson, N., Summers, B., 2002. Trade credit terms offered by small firms: Survey and
Monetary Policy. The University of Chicago Press, Chicago. empirical evidence. Journal of Business, Finance and Accounting 29, 317–351.