Journal of Banking & Finance 36 (2012) 28682883

Contents lists available at ScienceDirect

Journal of Banking & Finance

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

Trade credit, cash holdings, and nancial deepening:

Evidence from a transitional economy
Wenfeng Wu a, Oliver M. Rui b, Chongfeng Wu a,
a
b

Antai College of Economics and Management, Shanghai Jiao Tong University, Shanghai, China
China Europe International Business School 699 Hongfeng Road, Pudong Shanghai, China

a r t i c l e

i n f o

Article history:
Available online 24 April 2011
JEL classication:
G31
G32
Keywords:
Cash
Trade credit
Financial deepening
Receivables
Payables

a b s t r a c t
This paper investigates the effect of nancial deepening on the relationship between trade credit and cash
holdings among Chinese listed rms. We rst document an asymmetric effect of trade payables and
receivables on cash holdings, in that rms hold an additional $0.71 of cash for every $1 of credit payable
but use $1 of receivables as a substitute for only $0.15 of cash. We then nd that rms in regions with
higher levels of nancial deepening hold less cash for payables while substituting more receivables for
cash. A more highly developed nancial sector helps rms to better use trade credit as a short-term
nancing instrument. Finally, we nd that the ratio at which receivables are substituted for cash
increased following the implementation of the new receivables pledge policy in 2007, which allowed
rms to use receivables as security for loans. This policy event represents an exogenous shock that mitigates the endogeneity concern.
2011 Elsevier B.V. All rights reserved.

1. Introduction
Previous studies have found that trade credit constitutes a large
proportion of total assets. Rajan and Zingales (1995) nd that the
ratio of aggregate trade credit to total assets was 17.8% for US rms
in the early 1990s. Bartholdy and Mateus (2008) show that this ratio ranges between 16% and 24% across sixteen European countries.
According to the literature, rms use trade credit as a substitute
form of short-term nance to conventional institutional loans,
especially if they have been denied access to the institutional loan
market (Petersen and Rajan, 1997; Fishman and Love, 2003).1
Although rms can delay payment to their suppliers through
trade credit, they still need to hold some cash for forthcoming
trade credit obligations. Late trade credit payments have costs,
such as the cost of forgoing a possible cash discount, the possibility
of incurring late payment penalties, the opportunity cost associated with a possible deterioration in credit reputation, and a possible increase in the selling price set by the seller. At the same time,
rms often take on different roles in trade credit transactions.
Corresponding author. Tel.: +86 21 5230 1194; fax: +86 21 5230 1087.
E-mail addresses: wfwu@sjtu.edu.cn (W. Wu), oliver@baf.msmail.cuhk.edu.hk
(O.M. Rui), cfwu@sjtu.edu.cn (C. Wu).
1
Giannetti et al. (2011) provide a review of these theories. Some theories
emphasize operations-oriented motives such as informational advantage, price
discrimination, switching costs, product quality guarantees, and the protability
problem. Other theories focus on nancial motives including the collateral hypothesis
and the repayment enforcement hypothesis.
0378-4266/$ - see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankn.2011.04.009

Many companies, particularly those at intermediate points in the

value chain, use trade credit as a customer and provide it as a supplier. As a supplier of trade credit, a rm can accept credit receivables as a cash substitute by, for example, factoring receivables or
using them to secure loans. Consequently, trade payables and
receivables both have effects on rms cash holdings: payables increase cash holdings, while receivables have the opposite effect.
As cash is not productive, rms prefer to hold less cash for payables and expect to be able to substitute more receivables for cash.
The sensitivity of cash holdings to payables and receivables is affected by many factors including the nature of the rms payables
and receivables and the rm characteristics of suppliers and customers. In this study, we focus on another signicant factor: the
institutional nance environment. More specically, we investigate how nancial deepening affects the sensitivity of cash holdings to trade credit.
In the strand of literature that documents the importance of the
nancial system to economic growth (Dornbusch and Reynoso,
1989; Hasan et al., 2009), development of the nancial sector is
seen to improve the performance of nancial intermediaries,
which in turn provides the industrial sector with better nancial
services. Firms can consequently gain access to nance more easily
and at a lower cost. This leads to a lowering of the cost of cash
shortages for trade payables. In addition, a deeper nancial sector
increases the substitute ratio of receivables for cash, as rms can
factor receivables more easily or use them to secure loans from
nancial intermediaries at a lower cost. We thus hypothesize that

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

rms in regions with higher levels of nancial deepening hold less

cash to cover trade payables and have a higher substitute ratio of
receivables for cash.
Like their counterparts in other countries, Chinese rms exhibit
high ratios of trade credit to assets. Trade payables and receivables
represented 11% and 15%, respectively, of the total assets of Chinese listed rms over the period from 1999 to 2009. There are also
great variations in institutional quality across Chinese regions (Jin
et al., 2005), and the levels of nancial deepening across regions
are uneven. These characteristics make China a natural laboratory
for a cross-sectional investigation of whether and how nancial
sector development shapes the relationship between trade credit
and cash holdings. In 2007, China promulgated a new receivables
pledge policy that ofcially allowed rms to use receivables as collateral for bank borrowings. This policy event represents an exogenous shock that affords us the opportunity to examine the
inuence of nancial deepening on the sensitivity of cash holdings
to trade credit and mitigates concerns over endogeneity.
Using data from Chinese listed rms over the period from 1999
to 2009, we nd that rms need to hold an additional $0.71 of cash
for every $1 of credit payable, whereas $1 of credit receivable substitutes for only $0.15 of cash. This nding is not consistent with
the traditional wisdom that $1 of credit receivable covers $1 of
credit payable in cash. Firms with zero net trade credit (those for
which credit payable equals credit receivable) still need to hold
cash for payables. This asymmetric inuence of payables and
receivables on cash holdings suggests that past studies may have
drawn biased conclusions by treating net trade credit as just one
component of working capital in estimating its impact on cash
holdings. It is more appropriate to disentangle the impacts of payables and receivables on cash holdings, as they are separate and
different from each other.
We also nd that Chinese rms in regions with higher levels of
nancial deepening hold less cash to cover trade payables and have
a higher substitute ratio of credit receivables for cash. Additionally,
we nd that the new receivables pledge policy introduced in 2007
has exerted a considerable impact on the sensitivity of cash holdings to trade receivables, with the substitute ratio of receivables
for cash increasing signicantly following implementation of the
new policy. This effect is stronger in rms located in regions with
higher levels of nancial deepening. However, the new receivables
pledge policy does not affect the sensitivity of cash holdings to
trade payables, as it does not involve trade payables. Overall, these
results show that development of the nancial sector helps rms
to mitigate the asymmetric impact of trade payables and receivables on cash holdings. A higher level of nancial deepening improves the short-term nancing function of trade credit.
Furthermore, we analyze the inuence of related-party trades
and state ownership on the sensitivity of cash holdings to trade
credit. We nd that state-owned enterprises (SOEs) and rms that
have engaged in related-party trades hold less cash to cover payables and have a higher substitute ratio of receivables for cash. In
addition, related-party trades and state ownership have substitute
effects for nancial deepening on the relationship between trade
credit and cash holdings. The new receivables pledge policy has a
stronger effect on the sensitivity of cash holdings to receivables
in both rms that conduct related-party trades and SOEs, but its effect on the sensitivity of cash holdings to payables is not different
for related-party trade rms and state-owned entities. These results not only support our hypothesis, but also partially address
the concern that factors other than nancial deepening or the
new receivables pledge policy drive our results at the provinceyear level, as related-party trades and state ownership are rmspecic, but are not related to the province-year level.
This study contributes to the literature in several ways. First, it
extends research on the impact of nancial deepening (Dornbusch

2869

and Reynoso, 1989; Hasan et al., 2009). It investigates the inuence

of nancial deepening at the micro level by linking two important
rm operations: trade credit and cash holding policies.2 We demonstrate that nancial deepening can help rms to make better
use of trade credit as a short-term nancing instrument.
Second, we enrich existing studies on trade credit. Most of the
literature on trade credit focuses on why rms extend and take
credit (Petersen and Rajan, 1997; Cunat, 2007; Giannetti et al.,
2011). This study instead examines how trade credit inuences
the rms operations or, more specically, its cash management
policy. Our nding that payables and receivables exert different
impacts on cash holdings deepens our understanding of the difference between the demand and supply sides of trade credit.
Third, this study complements research on the determinants of
cash holdings (Opler et al., 1999; Dittmar et al., 2003; Ozkan and
Ozkan, 2004; Dittmar and Mahrt-Smith, 2007; Guney et al.,
2007). Past studies have not examined trade credit as a separate
determinant of cash holdings (Opler et al., 1999; Dittmar and
Mahrt-Smith, 2007).3 However, our results show that trade credit
plays a signicant role in explaining cash holdings, as it increases
the explanatory power of the regression model of their determinants. Furthermore, the asymmetric inuence of credit payables
and credit receivables indicates that it is better to disentangle their
distinct effects in considering their overall impact on cash holdings.
Finally, we nd that the ratio at which receivables are substituted for cash has increased since implementation of the new
receivables pledge policy in 2007 allowing rms to use receivables
as security for loans, which represents a deepening of the nancing
environment. This policy event also represents an exogenous shock
that mitigates concerns over endogeneity, which are common in
corporate nance research (Li and Prabhala, 2005).
The remainder of the paper is organized as follows. Section 2
introduces the institutional background on bank lending in Chinese rms, the use of trade credit, and the 2007 receivables pledge
policy. Section 3 develops the hypotheses. Section 4 describes the
data, variables, and methodology. Section 5 presents the empirical
results and Section 6 concludes the paper.

2. Institutional background
2.1. Banking lending in Chinese rms
The Chinese nancial system is dominated by a large banking
sector. The big four state-owned banks, comprising the Industrial
and Commercial Bank of China (ICBC), Agricultural Bank of China
(ABC), Bank of China (BOC), and China Construction Bank (CCB),
account for more than two-thirds of total deposits and loans in
China.4 At the end of 2009, the big four banks had deposits of
30.06 trillion RMB and loans of 17.32 trillion RMB, while the other
13 joint-stock commercial banks had deposits of 10.95 trillion
RMB and loans of 8.19 trillion RMB (Peoples Bank of China, 2010).
2
Only a few studies investigate the inuence of nancial deepening at the micro
level, such as the way in which nancing development affects rm growth and
investment (Demirg-Kunt and Maksimovic, 1998; Beck et al., 2004).
3
Among the determinants of cash holdings examined in the literature, the term
most closely related to trade credit is net working capital, which is regarded as a
liquidity substitute. Net trade credit (accounts receivable minus accounts payable) is
just one component of net working capital (Opler et al., 1999; Dittmar and MahrtSmith, 2007).
4
In the mid-1990s, the Chinese Government promulgated the Commercial Bank
Law and established three policy banks to take over the policy-related lending of
the above big four banks. Thus, the big four banks were commercialized. In addition
to the big four banks, there are another 13 joint-stock commercial banks such as the
Bank of Communications, China CITIC Bank, Shenzhen Development Bank, and China
Merchants Bank. Other than these 17 banks, there were 143 city commercial banks,
43 rural commercial banks, and 196 rural cooperative banks in China at the end of
2009 (China Banking Regulatory Commission, 2010).

2870

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

The main source of external nancing in China is loans from the

banking sector. From 1990 to 2008, the average annual ratio of total
bank loans to GDP was 82.4%.
Following the establishment of the Shanghai and Shenzhen
stock exchanges in 1990 and 1991, respectively, Chinas stock market has grown at a remarkable rate. At the end of 2010, 2063 rms
were listed on the two exchanges with a total market value of 4
trillion USD, which is equivalent to 66.7% of Chinas GDP (China
Securities Regulatory Commission, 2011). However, the funds
raised on the stock market are trivial in comparison with the
amount provided in bank loans. From 1990 to 2008, the average
annual ratio of total funds raised from the equity market to GDP
was only 0.81%. During this time, the corporate bond market
lagged behind the development of the equity market. The average
annual ratio of total funds raised from the corporate bond market
to GDP was only 0.76% between 1990 and 2008.
Fig. 1 illustrates the ratio of total bank loans to GDP and the ratio
of total funds raised from the equity and corporate bond markets to
GDP from 1990 to 2008. The ratio of bank loans to GDP ratio can be
seen to uctuate dramatically over the period, rising from a low of
61.6% in 1995 to a peak of 108% in 2003. The ratio of funds raised
from securities markets to GDP also uctuated over time, climbing
from less than one half of a percent in 1990 to 5% in 2007. Comparing the two ratios, the banking sector clearly dominates the securities markets, with total bank loans amounting to around 75 times
the amount of funds raised from the securities markets.
2.2. Use of trade credit in Chinese rms
In addition to securing funding through banks and securities
markets, Chinese rms can also access nance through trade credit
(Ge and Qiu, 2007; Cull et al., 2009). Several studies document that
Chinese rms use and offer trade credit much like rms in developed countries. Using data from a survey conducted by the Chinese
Academy of Social Sciences (CASS) in 2000, Ge and Qiu (2007) report that the averages of accounts receivable and accounts payable
in their sample represent 13% and 14% of rms total assets, respectively, while the ratios of accounts receivable and payable to total
sales are 27% and 23%, respectively. Cull et al. (2009) show a similar ratio of accounts receivable to total sales using a large sample
of more than 100,000 large and medium-sized industrial rms
from 1998 to 2003. Even listed rms, which may nd it easier gaining access to nance, tend to use and extend as much trade credit
as unlisted rms. Based on a sample of non-nancial listed rms
from 1999 to 2009, we nd that the ratios of accounts receivable
and payable to total assets are 14.5% and 10.9%, respectively, while
the ratios of accounts receivable and payable to total sales are 34%
and 20%, respectively.

0.05

Markets

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

0.6

1998

0.01
1997

0.7
1996

0.02

1995

0.8

1994

0.03

1993

0.9

1992

0.04

1991

Markets / GDP

Loan

1990

Loan / GDP

1.1

Year

Fig. 1. Bank loans ratio and funds raised from securities markets ratio, 19902008.
The bank loans ratio (loans/GDP) is calculated as total bank loans divided by the
GDP of China for the year. The ratio of funds raised from securities markets
(markets/GDP) is calculated as the sum of funds raised from the stock market and
corporate bonds divided by the GDP in that year.

Studies have also compared SOEs and non-SOEs in China

according to how they use and offer trade credit. Cull et al.
(2009) document that the ratio of trade credit extended to total
sales for SOEs is 36.5%, twice that of domestic private rms
(18%). However, Ge and Qiu (2007) nd that non-SOEs use more
trade credit than SOEs.
2.3. The new receivables pledge policy since 2007
Before the China Property Rights Law was enacted in 2007, secured transactions in China were regulated by a variety of laws.
The General Principles of Civil Law, the Security Law, and the Land
Administration Law dealt separately with various aspects of property ownership. This not only created confusion, but also increased
the number of bureaucratic procedures that applied, making the
process of creating, registering and enforcing security interests
costly, time-consuming and uncertain.
Furthermore, before 2007, non-possessory security interests in
China were allowed only for equipment and motor vehicles under
Article 34 of the 1995 Security Law. This meant that trade receivables, as one class of movable property, could not be used as collateral. According to a joint report by the World Bank and the Peoples
Bank of China, only 4% of Chinas commercial loans were nanced
by movable assets (FIAS/IFC, 2007).
Chinas new Property Rights Law was passed on March 16, 2007
and came into effect on October 1 of that year. The new Property
Rights Law merged all of the laws that previously regulated secured transactions under a uniform code. More importantly, under
Article 223 of the new law, trade receivables constitute a form of
property right that can be pledged. This means that trade receivables can be used as collateral to secure loans.
In accordance with the Property Rights Law, in 2007 Chinas
central bank promulgated the Measures for the Registration of
Pledge Receivables, which set out the detailed implementation
rules for using trade receivables as collateral to secure loans.
At the same time, Chinas central bank set up an electronic
registration and public notice system to facilitate the registration
process. According to the Credit Reference Center of the Peoples
Bank of China, more than 2 trillion RMB in loans were secured
against receivables by small and medium-sized enterprises
between October 2007 and September 2009 (Peoples Bank of
China, 2009).
3. Hypothesis development
3.1. Trade credit and cash holdings
3.1.1. Theories explaining cash holdings
Previous studies have found that cash and cash equivalents comprise a large percentage of rms assets. Dittmar and
Mahrt-Smith (2007) nd that the sum of all cash and marketable
securities represented more than 11% of the total assets of large
publicly traded US rms in 2003. Ozkan and Ozkan (2004) show
that in the United Kingdom, the average ratio of cash and
equivalent items to total assets was 10.3% between 1984 and
1999. Similar patterns have been identied in other countries.
For example, the cash ratios in Japan, Germany, and France are
17%, 9%, and 12%, respectively (Guney et al., 2007).
Several studies have attempted to explain why rms hold so
much cash, as the returns associated with cash are lower than
those for productive investment. Bates et al. (2009) list four
reasons rms hold more cash: the agency motive, the transaction
motive, the precautionary motive, and the tax motive. The agency
motive reects the argument that managers keep more free
cash ow in rms for their own private benet, such as empire

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

building or the consumption of more discretionary perquisites

(Dittmar et al., 2003; Guney et al., 2007; Harford et al., 2008).
The transaction cost motive suggests that rms hold cash because
converting assets into cash entails transaction costs (Opler et al.,
1999). The precautionary motive asserts that rms hold cash to
avoid future shortages of cash for investment, because raising
external nance is more costly than using internally generated
funds in the presence of asymmetric information (Opler et al.,
1999).5
3.1.2. Trade credit and cash holdings
The advantage of using trade credit is that the buyer need not
pay for goods on delivery and can enjoy a short deferment period
before payment is due. The seller of the goods extends credit to
the buyer, who is not required to deliver cash to the seller during
the period for which credit is extended. Thus, trade credit can be
regarded as a short-term nancing instrument. Outstanding accounts payable can create costs when cash is short. There is an
opportunity cost if a rm forgoes a cash discount and instead pays
the bill on the nal due date of the net period. If a rm postpones
payment beyond the net period due to a cash shortage, it will incur one of several possible costs of stretching accounts payable,
including late payment penalties or interest, a possible deterioration in its credit rating, and the cost of the cash discount forgone,
if any. Consequently, from the perspective of the precautionary
motive, rms that use trade credit must hold some additional
cash to meet their repayment obligations under the arrangements
for that credit to enjoy a cash discount and avoid late payment
penalties.
When a supplier offers trade credit, it does not receive cash at
the time it delivers the goods or services to the buyer. Rather,
the cash owing becomes an account or note receivable on the balance sheet. However, the supplier expects to collect cash from
these receivables at some point in the future, or can factor them
(use them as collateral for nance from a bank). These receivables
can thus be regarded as a cash substitute. This means that credit
receivables decrease cash holdings, in that the more receivables a
rm has, the less cash it holds.
3.1.3. Asymmetric effect of payables and receivables on cash holdings
Firms usually grant and receive trade credit at the same
time. As receivables can be regarded as a cash substitute, they
can be used to cover payables. This means that rms can hold
less cash to meet their payable obligations when they have
receivables on the balance sheet. However, risk aversion means
that $1 of receivables does not usually cover $1 of payables.
There is uncertainty about collecting receivables on time, and
rms are not always successful in collecting all receivables in
full. Some receivables are treated as bad debt losses over time.
Hence, receivables are discounted as a cash substitute, and are
not perfect substitutes for cash. Cash provides unconditional
liquidity for rms, whereas receivables are less liquid and subject
to credit risk.
Firms incur many costs associated with the late payment of
payables. Some rms prefer to pay payables earlier to obtain a discount. To guarantee their operations, rms generally hold enough
cash to cover their payables. If a rm uses receivables rather than
cash to pay its payables, then its creditors will ask for receivables
with a face value higher than the cash amount owed. Thus, from
the cash holding perspective, payables and receivables have an
asymmetric effect on cash holdings in that $1 of receivables does
not cover $1 of payables.
5
The tax motive theory argues that US rms that incur tax costs associated with
repatriating foreign earnings hold more cash (Foley et al., 2007).

2871

3.2. Financial deepening and the relationship between trade credit and
cash holdings
Although rms hold additional cash to cover their trade payables, the amount of cash they hold will not be the same as
the value of their payables. This is because cash, as a liquid asset,
has a lower rate of return due to the liquidity premium. Previous
studies have found that the value of $1 of cash is less than $1
(Dittmar and Mahrt-Smith, 2007). As a result, rms hold as little
cash as possible. The optimal amount of additional cash held to
repay trade credit reects a tradeoff between the lower rate of
return on cash and the cost of a cash shortage in paying for
credit.
We argue that the development of the nancial sector can
inuence the cost of cash shortages and thus has an impact on
the relationship between payables and cash holdings. A high level of nancial deepening provides the industrial sector with better nancial services, including easier access to nance, shorter
processing times to obtain funds, and lower nancing costs.
These benets help rms to reduce the cost of cash shortages
in paying for credit. Firms in regions with higher levels of nancial deepening are thus likely to hold less cash to cover their
trade payables.
In a more developed nancial sector, nancial intermediaries,
particularly banks, are better able to identify and pool the credit
risk of receivables and thus reduce the transaction costs incurred
in factoring receivables or using receivables to secure loans. The
better nancial services available in more developed nancial sectors lessen the costs incurred by rms in converting their receivables into cash. Hence, rms in regions with higher levels of
nancial deepening can substitute more of their receivables for
cash. Based on this discussion, we present the following
hypothesis.
Hypothesis 1. Firms in regions with higher levels of nancial
deepening hold less cash for payables. Firms in regions with higher
levels of nancial deepening can substitute more of their trade
receivables for cash.
3.3. Inuence of the new receivables pledge policy on the sensitivity of
cash holdings to trade credit
Trade receivables are an important part of rms overall assets,
representing 14.5% of the total assets of Chinese non-nancial
listed rms from 1999 to 2009. Before the new Property Law came
into effect in China in 2007, trade receivables could not be used as
collateral for securing loans. This affected the liquidity of receivables, which were classied as a liquid asset. Consequently, the
ability to convert receivables into cash was greatly reduced when
rms faced cash shortages. In other words, the substitute ratio of
receivables for cash was low.
However, Chinas new Property Rights Law allows trade receivables to be used as collateral to secure loans. According to the new
law, rms can factor their receivables when they face a cash shortage. This will enhance the short-term nancing function of trade
receivables. We thus predict that the new receivables pledge policy
has increased the substitute ratio of trade receivables for cash.
However, as this new policy has nothing to do with trade payables,
it does not affect the relationship between payables and cash holdings. We present the hypothesis as follows.
Hypothesis 2. The substitute ratio of trade receivables for cash has
increased since the new receivables pledge policy came into force
in 2007. The new receivables pledge policy does not affect the
sensitivity of cash holdings to trade payables.

2872

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

4. Research design
4.1. Data and sample
Our data are taken from the China Stock Market and Accounting
Research (CSMAR) database. We start with a sample of 1729 rms
and 14,313 rm-year observations from 1999 to 2009 for nonnancial A-share listed rms in China. After deleting any rm-year
observations with missing data or with either zero or negative total
assets or sales, our nal sample is reduced to 1626 rms and
13,229 rm-year observations.
4.2. Measurement of the key variables
4.2.1. Cash holding level
Following previous studies (Opler et al., 1999; Dittmar et al.,
2003; Dittmar and Mahrt-Smith, 2007), we dene the cash ratio
as the ratio of cash and cash equivalents to net assets, where net
assets are computed as total assets less cash and cash equivalents.
4.2.2. Trade credit
On the balance sheets of Chinese listed rms, accounts receivable and notes receivable are terms that describe situations in
which suppliers extend trade credit, whereas accounts payable
and notes payable are terms that describe situations in which customers receive trade credit. The variables CRDT_REV and CRDT_PAY
are respectively dened as the sum of accounts receivable and
notes receivable and the sum of accounts payable and notes payable, deated by net assets.6
4.2.3. Financial deepening
As discussed in Section 2.1, bank loans are the major source of
enterprise nance in China. Therefore, we use the ratio of total
bank loans to the GDP of the province in which the rm is located
to measure nancial deepening. Bank loan data are obtained from
the relevant annual issues of the Almanac of Chinas Finance and
Banking (ACFB). GDP data are taken from the relevant annual issues of the China Statistical Yearbook. Appendix A shows the average nancial deepening ratio from 1998 to 2008 by province.
4.3. Regression model and control variables
We extend the analysis of Opler et al. (1999) to trade credit
receivables and payables and use the following regression model
to calculate their inuence on cash holdings.

CASHi;t a b1 TRADECREDIT i;t b2 LIQUIDi;t b3 SIZEi;t1

b4 LEV i;t1 b5 DEBTMi;t1 b6 M=Bi;t1 b7 CAPEX i;t1
b8 CASHFLOW i;t1 b9 DIVIDENDi;t1 b10 TOP1i;t1
b11 STATEi;t1 b12 DEEPENi;t1
INDUSTRY; YEAR and PROVINCE Dummiesi;t1 ei;t
1
The dependent variable, CASH, is rm is cash holding ratio at time t.
The independent variables include trade credit variables and several
control variables. Based on prior studies (Opler et al., 1999; Dittmar
and Mahrt-Smith, 2007), we include the control variables of net
working capital ratio (LIQUID), rm size (SIZE), nancial leverage
(LEV), debt maturity (DEBTM), market-to-book ratio (M/B), capital
expenditure (CAPEX), cash ow (CASHFLOW), a dividend dummy
(DIVIDEND), the stake of the largest shareholder (TOP1), a stateowned enterprise dummy (STATE), and the nancial deepening
6
We also use total assets and total sales (or the cost of goods sold) as the deator,
but our results remain similar.

measure (DEEPEN). Except for the trade credit and net working capital variables, the other control variables are all calculated at the
beginning of the year to mitigate endogeneity problems. The denitions of these variables are discussed in the following section and
are summarized in Table 1.
The net working capital ratio (LIQUID) is a proxy for liquid assets and is dened as the ratio of net working capital (working capital minus cash and cash equivalents) to net assets. Net working
capital can be seen as a substitute for cash holdings, because rms
can use their liquid assets when they experience cash shortfalls.
There is a negative association between a rms cash holdings
and its liquid assets. Additionally, net trade credit (the sum of accounts and notes receivable minus the sum of accounts and notes
payable) is just one component of net working capital. To avoid
duplication in measuring trade credit, we also use an alternative
liquidity measure, LIQUID2, dened as the ratio of net working capital minus net trade credit to net assets.
Firm size (SIZE), dened as the natural logarithm of assets, is
known to be negatively associated with cash holdings. Larger rms
hold less cash, as they are more likely to be diversied and thus
less likely to experience nancial distress. They also face fewer
borrowing constraints and lower external nancing costs (Opler
et al., 1999; Dittmar et al., 2003). Leverage (LEV, total debt to total
assets) also exerts a negative impact on cash holdings, as higher
leverage indicates better access to external funds and reduces the
free cash ow problem (Opler et al., 1999; Harford et al., 2008).
Debt maturity (DEBTM, long-term debt to total debt) is related to
liquidity risk. We expect debt maturity to be positively associated
with cash holdings, as rms with longer maturity debt will hold
more liquidity in case they cannot meet xed debt payments during economic recessions (Morris, 1992)
We use the market-to-book ratio (M/B) to proxy for growth
opportunities. M/B is dened as the ratio of the book value of total
assets minus the book value of equity plus the market value of
equity to the book value of assets.7 Previous studies have found that
rms with more growth opportunities hold more cash (Opler et al.,
1999; Dittmar et al., 2003).
Findings on the inuence of the ratio of capital expenditure to
net assets (CAPEX) on cash holdings are mixed. Opler et al.
(1999) nd a positive impact of capital expenditure on cash holdings, whereas Harford et al. (2008) nd a negative relationship. The
cash ow ratio (CASHFLOW) is dened as net cash ow from operations divided by net assets. Opler et al. (1999) and Harford et al.
(2008) nd that rms with larger cash ows are associated with
larger cash holdings, whereas Ozkan and Ozkan (2004) identify a
negative impact of cash ow on cash holdings. The dividend
dummy (DIVIDEND) equals one in years in which a rm paid a cash
dividend and zero otherwise. Findings on the impact of dividend
payouts on cash holdings are also mixed. Opler et al. (1999),
Dittmar et al. (2003), and Harford et al. (2008) nd a negative
relationship, whereas Ozkan and Ozkan (2004) document a positive relationship.
In addition to these nancial control variables, we also include
several ownership structure variables.8 TOP1 is the percentage of

7
We take into account the special split share structure in China whereby some
shares are non-tradable on the stock market. We set the market value of non-tradable
shares as their book value, because these shares are usually transferred at a price
benchmarked against their book value. Thus, M/B is calculated as the ratio of the
number of tradable shares multiplied by the market price plus the number of nontradable shares multiplied by the book value of equity per share plus the book value
of total debt to the book value of total assets. We also use the standard calculation of
M/B as a robustness check, but our conclusions remain the same.
8
In our sample, the average percentage of shares held by the CEO is only 0.1%. As
management ownership in China is negligible, we do not include a management
ownership variable in our models. We rerun the regressions including the percentage
of CEO shares and nd that the results remain qualitatively the same.

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

2873

Table 1
Denitions of the variables.
Code

Denition

CASH
CRDT_PAY
CRDT_REV
LIQUID
LIQUID2

The ratio of cash and cash equivalents to net assets, where net assets are computed as total assets less cash and cash equivalents
The sum of accounts payable and notes payable, deated by net assets
The sum of accounts receivable and notes receivable, deated by net assets
The ratio of net working capital (working capital minus cash and cash equivalents) to net assets
The ratio of net working capital minus net trade credit to net assets, where net trade credit is dened as the sum of accounts receivable and notes
receivable minus the sum of accounts payable and notes payable
Firm size, calculated as the natural log of total assets
Financial leverage, calculated as total debt divided by total assets
The ratio of long-term debt to total debt
The ratio of the book value of total assets minus the book value of equity plus the market value of equity to the book value of assets
Capital expenditure, calculated as capital expenditure divided by total sales
The ratio of net cash ow from operations to net assets
A dummy that equals one for years in which the rm paid a cash dividend and zero otherwise
The percentage of shares held by the largest shareholder
A dummy variable that equals one if the rm is ultimately controlled by the government and zero otherwise
The ratio of bank loans to GDP in the province in which the rm is located
A dummy variable that equals one if the year is after 2007 and zero otherwise
A dummy variable that equals one if the rm has engaged in a related-party trade as a buyer in the year, and zero otherwise
A dummy variable that equals one if the rm has engaged in a related-party trade as the seller in the year, and zero otherwise

SIZE
LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN
POLICY
DRPT_BUY
DRPT_SELL

shares held by the largest shareholder.9 STATE is a dummy that

equals one if the rm is ultimately controlled by the government
and zero otherwise.10 Industry and year dummy variables are
included in the model to account for time-invariant industry heterogeneity and time trends. Industry classication is based on that of
the China Securities Regulatory Commission (CSRC), which recognizes 21 industries, with a one-digit code for non-manufacturing
industries and a two-digit code for manufacturing industries. A province dummy is added to capture differences in economic development, legal institutions (enforcement and judicial efciency), and
corruption across regions.
4.4. Regression method
We derive our main results using ordinary least squares (OLS)
regression. However, there is possible endogeneity between trade
credit and cash holdings, as a rms cash policy may have an impact on its trade credit policies. We thus conduct a two-stage
instrumental variables regression to correct for this endogeneity
problem. The rst stage of the procedure involves an OLS analysis
in which trade payables (or receivables) are regressed against the
same controls used for the OLS regressions and panel analyses plus
four other variables known to affect trade credit. These variables
are xed assets to total assets, the natural logarithm of rm age
in years (Petersen and Rajan, 1997; Giannetti et al., 2011), the
lagged cash holding ratio, and a dummy that indicates whether
the rm has engaged in related-party trade in that year. The estimate of trade payables (receivables) generated in the rst stage
is then included in the second-stage regression in which the
dependent variable is the cash holdings measure.
5. Empirical results
5.1. Descriptive statistics
Table 2 presents the descriptive statistics for the variables. We
winsorize the nancial variables at the 1% level to mitigate the
9
We report only the results for the model without the square term of TOP1. When
we include the square term of TOP1 in the regression model, it is not signicant and
our conclusions do not change.
10
We adopt the ownership classication scheme based on the identity of the
ultimate owner used in recent studies of Chinese listed rms (Wu et al., forthcoming).
State rms are ultimately controlled by the central and local governments or an
associated entity, such as central and local state-owned asset management bureaus,
the Ministry of Finance, or local nance bureaus.

effect of outliers.11 The mean value of CASH is 21.1%, from which we

calculate the ratio of cash and cash equivalents to total assets to be
17.4%. This suggests that listed rms in China hold as much cash as
rms in other countries. The ratio of trade credit payable (the sum of
accounts and notes payables) to net assets is 13.2%. This indicates
that in common with rms in other countries, Chinese listed rms
take a large amount of trade credit. The ratio of trade credit receivable (the sum of accounts and notes receivable) is 17.5%, which is
higher than the ratio of payables to net assets. This suggests that
listed rms extend more trade credit to customers than they receive
from suppliers.
We report a correlation analysis of the variables in Table 3. The
table shows that, except for correlations between LIQUID and LIQUID2, the correlations between the variables are not very high.
We also check the variance ination factors (VIFs) of the coefcients in our models and nd that they are less than 10. This suggests that multicollinearity is not a serious concern.
5.2. Asymmetric impact of trade payables and receivables on cash
holdings
Table 4 presents the results of multivariate regressions on the
relationship between trade credit and cash holdings. Panel A of Table 4 reports the results of the four regression models, and Panel B
reports the results of the tests of equality of the coefcients. LIQUID
and LIQUID2 (excluding the trade credit terms) are included in
Models (1) and (2), respectively, to proxy for substitute liquidity.
Model (3) separates LIQUID into two terms: LIQUID2 and net trade
credit extended (trade receivables minus trade payables). We include the trade credit variables in Model (4) to estimate the inuence of trade credit on cash holdings. The constant term, industry,
year, and province dummies are included in all of the regressions,
although the results are not reported for brevity. The p-values in
the panel regressions are based on standard errors corrected for
the clustering of rms (Petersen, 2009).
First, we compare the inuence of liquidity assets on cash holdings in Models (1)(3). The three measures of LIQUID, LIQUID2, and
the difference between LIQUID and LIQUID2 are negative and significant at less than the 1% signicance level. The signs and signicance levels of the other variables in Model (3) are the same as
those in Model (1). This suggests that replacing LIQUID with LIQUID2 does not affect the impact of the other variables on cash
11
We also perform a robustness check without winsorizing the variables. The
results remain qualitatively similar.

2874

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

Table 2
Descriptive Statistics for the Main Variables.
Variable

Mean

Standard deviation

Min

P25

Median

P75

Max

CASH
CRDT_PAY
CRDT_REV
LIQUID
LIQUID2
SIZE
LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN
DRPT_BUY
DRPT_SELL

13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229
13,229

0.211
0.132
0.175
0.088
0.131
21.063
0.521
0.146
1.555
0.059
0.068
0.507
0.398
0.687
1.010
0.453
0.474

0.223
0.116
0.136
0.293
0.285
1.143
0.281
0.174
0.655
0.061
0.117
0.500
0.167
0.464
0.369
0.498
0.499

0.003
0.000
0.001
1.512
1.528
10.391
0.074
0
0.950
0.000
0.488
0
0.004
0
0.447
0
0

0.077
0.048
0.065
0.216
0.245
20.356
0.359
0.010
1.140
0.013
0.038
0
0.266
0
0.773
0
0

0.146
0.099
0.148
0.063
0.107
20.960
0.500
0.077
1.340
0.039
0.070
1
0.378
1
0.878
0
0

0.264
0.180
0.253
0.085
0.023
21.693
0.633
0.222
1.713
0.082
0.115
1
0.526
1
1.172
1
1

1.718
0.575
0.612
0.503
0.493
27.940
2.160
0.726
4.821
0.296
0.464
1
0.886
1
2.176
1
1

This table presents the summary statistics for the variables. The denitions of the variables are as presented in Table 1.

holdings. However, as shown in Model (3), the coefcient of LIQUID2 is 0.28 and its absolute magnitude is lower than that of
the coefcient of net trade credit extended (LIQUIDLIQUID2),
which is 0.32. The results of the equality tests reported in Panel
B also show that these two coefcients are statistically different,
which suggests that the impact of net trade credit is different from
that of the other components of net working capital. This implies
that it is necessary to separate trade credit terms from net working
capital when considering their impact on cash holdings.
In Model (4), we include CRDT_PAY and CRDT_REV to investigate
their individual inuence on cash holdings. The coefcient of
CRDT_PAY is 0.71 and is statistically signicant. Because the
denominators of the CRDT_PAY and CASH ratios are the same, the
coefcient of CRDT_PAY of 0.71 indicates that rms hold an additional $0.71 of cash for every $1 of trade credit payable. The results
of the equality tests documented in Panel B show that the coefcient value of $0.71 is signicantly less than 1. This suggests that
rms do not hold the same amount of cash for the precautionary
repayment of trade credit.
The coefcient of CRDT_REV is statistically signicantly negative. This indicates that rms treat receivables as cash substitutes
and reduce their cash holdings accordingly. However, the coefcient is only 0.15. This means that $1 of trade credit receivable
substitutes for only $0.15 of cash, which is far less than the $0.71
of cash needed for every $1 of trade credit payable. The results of
the tests on the equality of the coefcients in Panel B show that
the sum of these two coefcients is signicantly different from
zero. This clearly demonstrates that $1 of receivables is not equivalent to $1 of payables in terms of cash holdings.
In an unreported regression, we replicate Table 4 using the
FamaMacBeth model. A cross-sectional regression is estimated
to eliminate the problem of serial correlation in the residuals of a
time-series cross-sectional regression. The results remain unchanged, lending additional support for the asymmetric impact
of credit payable and receivable on cash holdings and suggesting
that rms with zero net trade credit still need to hold some cash
for payables. However, if we treat trade credit terms merely as
components of working capital, as is the standard approach in
the literature, we might mistakenly conclude that no cash is
needed for zero net trade credit, as receivables cover payables.
Clearly, it is essential to treat trade credit payable and trade credit
receivable differently when investigating their respective impact
on cash holdings.
Table 4 shows that rm size (SIZE), nancial leverage (LEV), and
capital expenditure (CAPEX) are negatively associated with cash

holdings. This indicates that rms that are smaller and have lower
nancial leverage and less capital expenditure have larger cash
holdings. The coefcient of DEBTM is signicantly positive, which
suggests that rms with longer-term debt hold more cash to avoid
the risk of unexpected liquidity problems (Morris, 1992). The signicantly positive coefcient of CASHFLOW indicates that rms
with more net cash ow from operations hold more cash (Opler
et al., 1999; Dittmar et al., 2003). The coefcient of DIVIDEND is signicantly positive, which is consistent with the nding of Ozkan
and Ozkan (2004) that dividend-paying rms hold more cash to
avoid running out of funds to meet their dividend payments. In
terms of ownership characteristics, TOP1 is insignicant. This indicates that the two conicting forces exerted by largest block ownership on cash holdings may cancel each other out. The coefcient
of STATE in Model (4) is signicantly negative, suggesting that
state-owned enterprises hold less cash due to their advantages in
obtaining nancing. The coefcients of the nancial deepening variable are also insignicant.
5.3. Financial deepening and the relation between trade credit and
cash holdings
Table 5 reports the results of the regression of the impact of
nancial deepening on the relationship between trade credit
and cash holdings. Models (1) and (2) present the results for
the subsamples based on nancial deepening. Model (1) shows
the results for the subsample with a lower level of nancial deepening and Model (2) those for the subsample with a higher level
of nancial deepening. Models (3) and (4) present the results of
the regression of the interaction terms between the nancial
deepening variable and credit payables and credit receivables,
respectively. Model (5) includes the interaction terms of the
nancial deepening variable with credit payables and credit
receivables.
As shown in Model (1) of Table 5, for the subsample with a lower level of nancial deepening, the coefcient of CRDT_PAY is 0.720,
whereas that in Model (2) for the subsample with a higher level of
nancial deepening is 0.709. This suggests that rms located in regions with greater nancial depth hold less precautionary cash for
payables. As greater nancial depth enables rms to borrow from
banks, it reduces the probability that they will experience a cash
shortage when paying off their payables. The coefcient of
CRDT_REV in Model (1) is 0.138, whereas in Model (2) it is
0.199. This shows that the substitute ratio of receivables for cash
in rms in regions with a higher level of nancial deepening is

Table 3
Pearson correlation matrix for the variables.
Variable

CASH

CRDT_PAY

0.139***
(0.001)

CRDT_REV

0.048***
(0.001)

CRDT_PAY

CRDT_REV

LIQUID

LIQUID2

SIZE

LEV

DEBTM

0.046***
(0.001)

0.102***
(0.001)

0.147***
(0.001)

0.855***
(0.001)

SIZE

0.184***
(0.001)

0.096***
(0.001)

0.222***
(0.001)

0.066***
(0.001)

0.214***
(0.001)

LEV

0.266***
(0.001)

0.242***
(0.001)

0.063***
(0.001)

0.692***
(0.001)

0.649***
(0.001)

0.038***
(0.001)

DEBTM

0.159***
(0.001)

0.281***
(0.001)

0.284***
(0.001)

0.032***
(0.001)

0.052***
(0.001)

0.296***
(0.001)

0.125***
(0.001)

0.118***
(0.001)

0.101***
(0.001)

0.030***
(0.001)

0.047***
(0.001)

0.111***
(0.001)

0.387***
(0.001)

0.049***
(0.001)

0.081***
(0.001)

0.041***
(0.001)

0.097***
(0.001)

0.234***
(0.001)

0.027***
(0.001)

0.046***
(0.001)

0.187***
(0.001)

0.158***
(0.001)

0.255***
(0.001)

CASHFLOW

0.339***
(0.001)

0.039***
(0.001)

0.161***
(0.001)

0.265***
(0.001)

0.331***
(0.001)

0.157***
(0.001)

0.482***
(0.001)

0.016*
(0.072)

DIVIDEND

0.175***
(0.001)

0.025***
(0.004)

0.140***
(0.001)

0.138***
(0.001)

0.221***
(0.001)

0.292***
(0.001)

0.251***
(0.001)

TOP1

0.028***
(0.001)

0.008
(0.334)

0.007
(0.424)

0.115***
(0.001)

0.114***
(0.001)

0.206***
(0.001)

STATE

0.036***
(0.001)

0.005
(0.542)

0.060***
(0.001)

0.020**
(0.018)

0.052***
(0.001)

0.124***
(0.001)

0.021**
(0.014)

0.005
(0.587)

0.036***
(0.001)

0.026***
(0.002)

LIQUID2

M/B
CAPEX

DEEPEN

CASHFLOW

DIVIDEND

TOP1

STATE

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

0.204***
(0.001)

0.001
(0.901)

CAPEX

0.275***
(0.001)
0.170***
(0.001)

LIQUID

M/B

0.012
(0.160)
0.054***
(0.001)

0.239***
(0.001)

0.093***
(0.001)

0.081***
(0.001)

0.270***
(0.001)

0.341***
(0.001)

0.159***
(0.001)

0.044***
(0.001)

0.149***
(0.001)

0.094***
(0.001)

0.164***
(0.001)

0.177***
(0.001)

0.238***
(0.001)

0.094***
(0.001)

0.090***
(0.001)

0.143***
(0.001)

0.045***
(0.001)

0.066***
(0.001)

0.101***
(0.001)

0.303***
(0.001)

0.064***
(0.001)

0.033***
(0.001)

0.028***
(0.001)

0.002
(0.799)

0.035***
(0.001)

0.022**
(0.010)

0.073***
(0.001)

0.087***
(0.001)

0.046***
(0.001)

This table reports the Pearson correlation matrix for the variables. The p-values are presented in parentheses below the correlation coefcients, where the denitions of the variables are as presented in Table 1.
*
Signicance at the 10% level.
**
Signicance at the 5% level.
***
Signicance at the 1% level.

2875

2876

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

relationship between cash holdings and credit payables and that

between cash holdings and receivables.

Table 4
Multivariate regression of the impact of trade credit on cash holdings.
Model 1
Panel A: Regression results
LIQUID
0.319***
(0.000)
LIQUID2

Model 2

LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN

0.201***
(0.000)

0.046***
(0.000)
0.350***
(0.000)
0.078***
(0.000)
0.007**
(0.041)
0.521***
(0.000)
0.535***
(0.000)
0.054***
(0.000)
0.013
(0.220)
0.007*
(0.079)
0.031*
(0.068)

0.044***
(0.000)
0.232***
(0.000)
0.021*
(0.059)
0.012***
(0.000)
0.384***
(0.000)
0.593***
(0.000)
0.057***
(0.000)
0.012
(0.268)
0.000
(0.927)
0.021
(0.167)

0.284***
(0.000)
0.320***
(0.000)
0.047***
(0.000)
0.327***
(0.000)
0.067***
(0.000)
0.007**
(0.029)
0.507***
(0.000)
0.530***
(0.000)
0.054***
(0.000)
0.011
(0.297)
0.006
(0.118)
0.030*
(0.075)

13,229
0.349

13,229
0.310

13,229
0.338

CRDT_PAY
CRDT_REV
Sample size
Adj-R2

Model 4

5.4. Inuence of the new receivables pledge policy

LIQUIDLIQUID2
SIZE

Model 3

F-value
Panel B: Test of equality of the coefcients (F-statistics)
LIQUID2 = LIQUIDLIQUID2
25.16***
CRDT_PAY = 1
398.30***
CRDT_PAY + CRDT_REV = 0
623.33***

0.360***
(0.000)

0.048***
(0.000)
0.438***
(0.000)
0.188***
(0.000)
0.001
(0.704)
0.466***
(0.000)
0.510***
(0.000)
0.052***
(0.000)
0.010
(0.367)
0.007**
(0.046)
0.019
(0.205)
0.714***
(0.000)
0.153***
(0.000)
13,229
0.374
p-Value
0.000
0.000
0.000

This table reports the results of the regression of the association between the trade
credit variables and cash holdings. Panel A presents the regression results and Panel
B shows the results of the test of equality of the coefcients. The constant term,
industry dummies, year dummies, and province dummies are included in the
regression but are not reported. The p-values, adjusted for clustering at the rm
level, are presented in parentheses below the estimates, where the denitions of
the variables are as presented in Table 1.
*
Signicance at the 10% level.
**
Signicance at the 5% level.
***
Signicance at the 1% level.

higher than that in rms in regions with less nancial depth. These
results are consistent with our rst hypothesis.
To test whether the coefcient of trade credit differs across
rms located in regions with different levels of nancial depth,
we include the interaction terms between the nancial deepening
variable (DEEPEN) and the trade credit variables in the model. The
columns for Models (3)(5) in Table 5 report the results. The interaction terms between DEEPEN and CRDT_PAY are signicantly negative in Models (3) and (4), whereas those between DEEPEN and
CRDT_REV are both signicantly negative across Models (4) and
(5). This supports our rst hypothesis, that is, nancial deepening
helps rms to reduce cash holdings for payables and to increase
the substitute ratio of receivables for cash.
Overall, these results suggest that rms in regions with higher
levels of nancial deepening are less likely to experience cash
shortages. Financial depth reduces the cost of a shortage of cash
to repay credit and increases the substitute ratio of receivables
for cash. Financial deepening thus negatively affects both the

Table 6 reports the results for the regression of the impact of the
new receivables pledge policy on the relationship between trade
credit and cash holdings. We divide the full sample into two subsamples comprising pre- and post-2007 observations, respectively,
as the receivables pledge policy was implemented in 2007. Model
(1) presents the results for the pre-2007 subsample and Model (2)
those for the post-2007 subsample. To test the statistical signicance of the impact of the receivables pledge policy, we include
the interaction term between the trade credit variables and a 1year dummy variable (POLICY) that takes the value of one if the
observation is after 2007, and zero otherwise. The Model (3)(5)
columns report the results of the regressions with the interaction
terms.
In comparing the coefcients of CRDT_PAY and CRDT_REV between Model (1) and Model (2) in Table 6, we nd that the coefcient of CRDT_PAY decreases a little after implementation of the
new receivables pledge policy, whereas the coefcient of CRDT_REV
decreases substantially after 2007, falling from 0.155 to 0.203.
This suggests that the new policy has substantially increased the
substitute ratio of receivables for cash, as it ofcially allows rms
to use receivables as collateral to borrow from banks.12 However,
as the new policy has nothing to do with payables, it does not significantly affect rms cash holdings for payables. These results are consistent with our second hypothesis.
Model (5) of Table 6, which includes both of the interaction
terms between CRDT_PAY and CRDT_REV and the policy year dummy (POLICY), may be better specied than Models (3) and (4),
which have only a single interaction term. Therefore, we mainly
derive our conclusion from the results of Model (5). In Model (5),
the coefcient of the interaction term between CRDT_PAY and POLICY is insignicant, whereas that of the interaction term between
CRDT_REV and POLICY is signicantly negative. These results support our second hypothesis, that is, the new receivables pledge policy helps rms to increase the substitute ratio of receivables for
cash, but has no signicant inuence on the relationship between
credit payables and cash holdings. Given that the policy represents
a deepening of the nancing environment, our nding also supports the view that nancial deepening has a positive impact on
the use of trade receivables as a cash substitute.13
Furthermore, we compare the inuence of the policy on rms
located in regions with different levels of nancial deepening. As
shown in Models (1) and (2) of Table 7, the coefcient of the interaction term between CRDT_PAY and DEEPEN decreases from
0.055 to 0.080 after implementation of the new receivables
pledge policy, whereas the coefcient of the interaction term between CRDT_REV and DEEPEN decreases substantially after 2007,
falling from 0.007 to 0.125. This suggests that the new receivables pledge policy has had a stronger impact on rms in regions
with higher levels of nancial deepening. This is a reasonable

12
As our sample is limited to publicly traded rms, which tend to nd it easier to
secure loans when needed, the new receivables pledge policy may be not be
important to them. Our results will be understated in relation to non-publicly traded
rms. Therefore, the signicant effect of the new policy on publicly traded rms
13
The global nancial crisis that struck after 2007 may also have presented rms
with severe liquidity problems, in that it may have reduced the substitute ratio of
receivables for cash and required rms to hold more cash for payables. Thus, the
nancial crisis is likely to have exerted a negative effect on the sensitivity of cash to
receivables, whereas the new receivables pledge policy probably exerted a positive
inuence on the same measure. Our results suggest that the effect of the pledge policy
dominates that of the nancial crisis.

2877

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

Table 5
Financial deepening and the relationship between trade credit and cash holdings.

LIQUID2
SIZE
LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN
CRDT_PAY
CRDT_REV

Model 1: Subsample with a lower level of nancial

deepening

Model 2: Subsample with a higher level of nancial

deepening

Model 3

Model 4

Model 5

0.294***
(0.000)
0.040***
(0.000)
0.418***
(0.000)
0.164***
(0.000)
0.006
(0.114)
0.416***
(0.000)
0.463***
(0.000)
0.056***
(0.000)
0.018
(0.186)
0.012**
(0.012)
0.030
(0.277)
0.720***
(0.000)
0.138***
(0.000)

0.428***
(0.000)
0.053***
(0.000)
0.452***
(0.000)
0.216***
(0.000)
0.001
(0.860)
0.564***
(0.000)
0.561***
(0.000)
0.051***
(0.000)
0.028*
(0.075)
0.008
(0.169)
0.002
(0.932)
0.709***
(0.000)
0.199***
(0.000)

0.359***
(0.000)
0.048***
(0.000)
0.439***
(0.000)
0.189***
(0.000)
0.001
(0.694)
0.465***
(0.000)
0.510***
(0.000)
0.052***
(0.000)
0.010
(0.351)
0.007**
(0.047)
0.028*
(0.065)
0.792***
(0.000)
0.142***
(0.000)
0.077**
(0.035)

0.359***
(0.000)
0.048***
(0.000)
0.438***
(0.000)
0.189***
(0.000)
0.001
(0.693)
0.464***
(0.000)
0.511***
(0.000)
0.052***
(0.000)
0.009
(0.382)
0.007*
(0.051)
0.029*
(0.058)
0.715***
(0.000)
0.076**
(0.034)

0.065**
(0.028)

0.359***
(0.000)
0.048***
(0.000)
0.438***
(0.000)
0.189***
(0.000)
0.001
(0.688)
0.464***
(0.000)
0.510***
(0.000)
0.052***
(0.000)
0.010
(0.366)
0.007*
(0.050)
0.035**
(0.030)
0.775***
(0.000)
0.090**
(0.014)
0.060**
(0.048)
0.050**
(0.037)

13,229
0.375

13,229
0.375

CRDT_PAY  DEEPEN
CRDT_REV  DEEPEN
Sample size
Adj-R2

6668
0.352

6561
0.389

13,229
0.375

This table reports the results of the regression of nancial deepening on the relationship between trade credit and cash holdings. The constant term, industry dummies, year
dummies, and province dummies are included in the regression but are not reported. The p-values, adjusted for clustering at the rm level, are presented in parentheses
below the estimates, where the denitions of the variables are as presented in Table 1.
*
Signicance at the 10% level.
**
Signicance at the 5% level.
***
Signicance at the 1% level.

result, as rms in regions with higher levels of nancial deepening

have access to better nancial services, meaning they are better
able to use receivables to factor or secure loans and thus benet
more from the new receivables pledge policy. The new receivables
pledge policy can be treated as a de jure measure of nancial
deepening, while the ratio of bank loans to GDP is a de facto
measure of nancial deepening. The two measures complement
each other.
We include the three-way interaction term among CRDT_PAY,
DEEPEN, and the POLICY dummy and the three-way interaction
term among CRDT_REV, DEEPEN, and POLICY in Models (3)(5) to
test the difference in the impact of the new receivables pledge
policy between rms located in regions with lower and higher levels of nancial deepening. We focus on the results of Model (5)
with the two three-way interaction terms. The coefcient of the
three-way interaction term among CRDT_PAY, DEEPEN, and POLICY
is insignicant, whereas that of the three-way interaction term
among CRDT_REV, DEEPEN, and POLICY is signicantly negative.
These results also support our conjecture that the impact of the
new receivables pledge policy on receivables is more pronounced
for rms in regions with higher levels of nancial deepening, but
the effect of the new policy on the payables of rms is no different
in regions with lower and higher levels of nancial deepening.
As DEEPEN is at the province-year level and POLICY is at the year
level, the above regressions may not rule out an alternative hypothesis that factors other than nancial deepening/pledge policy
change drive the results at the province-year level. To address this
concern, we employ two rm-specic variables, related-party

trades and state ownership, which affect the sensitivity of cash holdings to trade credit, but are not related to the province-year level.14
5.5. Inuence of related-party trades
The type of institution that is the counterparty of a trade credit
relationship may affect the sensitivity of cash holdings to payables
and receivables. Related parties are common counterparty types,
and can include shareholders, companies in the same group as
shareholders, and subsidiaries. Trades between listed rms and
these types of related parties are called related-party trades. The
close relationship between a rm and a related party reduces the
cost of delayed payment for payables and increases the probability
of receivables being converted into cash when needed.
We use two dummies to capture whether a rm has engaged in
related-party sales or purchases. The dummy DRPT_BUY equals one
if a rm has engaged in a related-party trade as a buyer and zero
otherwise. The dummy DRPT_SELL equals one if a rm has engaged
in a related-party trade as the seller and zero otherwise. We employ regression models with interaction terms between DRPT_BUY
and CRDT_PAY and between DRPT_SELL and CRDT_REV for the two
subsamples based on the level of nancial deepening.
To compare the difference in the interaction terms between
rms located in regions with lower and higher levels of nancial
deepening, we run a regression with three-way interaction terms
14
We thank an anonymous referee for helping us to point out and address this
concern.

2878

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

Table 6
Impact of the new receivables pledge policy on the relationship between trade credit and cash holdings.
Model 1: Pre-2007 subsample
LIQUID2
SIZE
LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN
CRDT_PAY
CRDT_REV

***

0.369
(0.000)
0.049***
(0.000)
0.435***
(0.000)
0.170***
(0.000)
0.010**
(0.013)
0.461***
(0.000)
0.477***
(0.000)
0.060***
(0.000)
0.028**
(0.017)
0.009**
(0.044)
0.063***
(0.000)
0.747***
(0.000)
0.155***
(0.000)

Model 2: Post-2007 subsample

***

0.357
(0.000)
0.046***
(0.000)
0.476***
(0.000)
0.211***
(0.000)
0.003
(0.529)
0.477***
(0.000)
0.538***
(0.000)
0.058***
(0.000)
0.020
(0.345)
0.006
(0.371)
0.064***
(0.000)
0.738***
(0.000)
0.203***
(0.000)

CRDT_PAY  POLICY

Model 3
***

0.361
(0.000)
0.048***
(0.000)
0.439***
(0.000)
0.188***
(0.000)
0.001
(0.656)
0.464***
(0.000)
0.510***
(0.000)
0.053***
(0.000)
0.009
(0.387)
0.007**
(0.047)
0.020
(0.174)
0.742***
(0.000)
0.142***
(0.000)
0.064**
(0.020)

CRDT_REV  POLICY
Sample size
Adj-R2

8892
0.348

4337
0.390

13,229
0.375

Model 4
***

Model 5

0.084***
(0.002)

0.360***
(0.000)
0.049***
(0.000)
0.440***
(0.000)
0.188***
(0.000)
0.002
(0.631)
0.463***
(0.000)
0.512***
(0.000)
0.053***
(0.000)
0.009
(0.388)
0.008**
(0.044)
0.020
(0.178)
0.735***
(0.000)
0.126***
(0.000)
0.034
(0.264)
0.069**
(0.025)

13,229
0.375

13,229
0.375

0.360
(0.000)
0.049***
(0.000)
0.440***
(0.000)
0.188***
(0.000)
0.001
(0.646)
0.463***
(0.000)
0.513***
(0.000)
0.053***
(0.000)
0.009
(0.380)
0.008**
(0.043)
0.019
(0.192)
0.722***
(0.000)
0.124***
(0.000)

In 2007, China adopted the Property Rights Law, which prescribes that accounts receivable can be used as collateral to borrow from banks. In accordance with the Property
Rights Law, Chinas central bank promulgated the Measures for the Registration of Pledge Receivables in the same year. This table reports the results of regressions testing for
the inuence of the new receivables pledge policy on the relationship between trade credit and cash holdings. The constant term, industry dummies, year dummies, and
province dummies are included in the regression but are not reported. The p-values, adjusted for clustering at the rm level, are presented in parentheses below the
estimates, where the denitions of the variables are as presented in Table 1.

Signicance at the 10% level.

**
Signicance at the 5% level.
***
Signicance at the 1% level.

among CRDT_PAY, DRPT_BUY, and DEEPEN, and among CRDT_REV,

DRPT_SELL, and DEEPEN. To capture the interactive effect of related-party trades and the new receivables pledge policy, we also
conduct a similar regression with three-way interaction terms
among CRDT_PAY, DRPT_BUY, and the POLICY year dummy, and
among CRDT_REV, DRPT_SELL, and POLICY. Table 8 reports the
regression results.
The results of Model (1) reported in Table 8 show that the coefcients of the interaction terms between DRPT_BUY and CRDT_PAY
and between DRPT_SELL and CRDT_REV are both signicantly negative. This indicates that in regions with lower levels of nancial
deepening, rms making related-party purchases tend to hold less
cash for payables, whereas the substitute ratio of receivables for
cash in rms making related-party sales tends to be higher. However, the results for Model (2) show that related-party trades do
not affect the sensitivity of cash holdings to trade credit in regions
with higher levels of nancial deepening. Thus, the effect of related-party trades is mainly driven by rms in regions with lower
levels of nancial deepening.
The results of Model (3) show that the three-way interaction
terms among DRPT_BUY, CRDT_PAY, and DEEPEN, and among
DRPT_SELL, CRDT_REV, and DEEPEN are both signicantly positive.
This indicates that related-party trades partially mitigate the problems imposed by lower levels of nancial deepening. As a close
relationship with a related party can enable a rm to collect receivables more easily and reduce the cost of delayed payment for payables when facing a cash shortage, rms that conduct related-party

trades do not need to keep as much cash for payables and can substitute more receivables for cash. In other words, the roles of related-party trades and nancial deepening are substitutable.
Model (4) in Table 8 shows the results for the interactive inuence of related-party trades and the new receivables pledge policy
on the sensitivity of cash holdings to trade credit. The coefcient of
the three-way interaction term among CRDT_PAY, DRPT_BUY, and
POLICY is not signicant, which indicates that the new receivables
pledge policy does not have a signicantly different effect on the
sensitivity of cash holdings to payables between rms that have
engaged in related-party purchases and those that have not. However, the three-way interaction term among CRDT_REV, DRPT_SELL,
and POLICY is signicantly negative. This indicates that the new
receivables pledge policy has a stronger effect on the sensitivity
of cash holdings to receivables for rms that have engaged in related-party sales. Just as the new receivables pledge policy is complementary to nancial deepening in regard to the sensitivity of
cash holdings to receivables, it is also complementary to relatedparty trades. These results also support our second hypothesis that
the new receivables pledge policy helps rms to increase the
substitute ratio of receivables for cash, but does not affect the sensitivity of cash holdings to trade payables.
5.6. Inuence of state ownership
During its gradual economic transition, China has maintained
extensive state control over the economy. Many government

2879

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

Table 7
Interaction of the impact of the new receivables pledge policy and nancial deepening on the relationship between trade credit and cash holdings.
Model 1: Pre-2007 subsample
LIQUID2
SIZE
LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN
CRDT_PAY
CRDT_REV
CRDT_PAY  DEEPEN
CRDT_REV  DEEPEN

***

0.374
(0.000)
0.049***
(0.000)
0.439***
(0.000)
0.176***
(0.000)
0.010***
(0.008)
0.458***
(0.000)
0.472***
(0.000)
0.056***
(0.000)
0.015
(0.203)
0.010**
(0.032)
0.073***
(0.000)
0.807***
(0.000)
0.145***
(0.001)
0.055
(0.243)
0.007
(0.844)

Model 2: Post-2007 subsample

***

0.358
(0.000)
0.046***
(0.000)
0.476***
(0.000)
0.201***
(0.000)
0.001
(0.778)
0.470***
(0.000)
0.542***
(0.000)
0.055***
(0.000)
0.017
(0.411)
0.000
(0.954)
0.043
(0.737)
0.806***
(0.000)
0.079
(0.319)
0.080
(0.264)
0.125*
(0.093)

CRDT_PAY  DEEPEN  POLICY

Model 3
***

0.360
(0.000)
0.048***
(0.000)
0.439***
(0.000)
0.189***
(0.000)
0.001
(0.675)
0.463***
(0.000)
0.510***
(0.000)
0.052***
(0.000)
0.009
(0.375)
0.007*
(0.052)
0.038**
(0.019)
0.779***
(0.000)
0.085**
(0.022)
0.043
(0.269)
0.054*
(0.098)
0.050**
(0.041)

CRDT_REV  DEEPEN  POLICY

Sample size
Adj-R2

8892
0.359

4337
0.396

13,229
0.375

Model 4
***

Model 5

0.070***
(0.004)

0.360***
(0.000)
0.049***
(0.000)
0.440***
(0.000)
0.189***
(0.000)
0.001
(0.659)
0.462***
(0.000)
0.512***
(0.000)
0.053***
(0.000)
0.009
(0.378)
0.007**
(0.048)
0.038**
(0.018)
0.771***
(0.000)
0.077**
(0.037)
0.042
(0.276)
0.049
(0.138)
0.017
(0.555)
0.061**
(0.035)

13,229
0.375

13,229
0.375

0.359
(0.000)
0.049***
(0.000)
0.440***
(0.000)
0.189***
(0.000)
0.001
(0.659)
0.462***
(0.000)
0.513***
(0.000)
0.053***
(0.000)
0.009
(0.375)
0.007**
(0.047)
0.037**
(0.020)
0.769***
(0.000)
0.078**
(0.036)
0.046
(0.226)
0.047
(0.152)

This table reports the results of regressions testing for the inuence of the new receivables pledge policy on the relationship between trade credit and cash holdings in regions
with different levels of nancial deepening. The constant term, industry dummies, year dummies, and province dummies are included in the regression but are not reported.
The p-values, adjusted for clustering at the rm level, are presented in parentheses below the estimates, where the denitions of the variables are as presented in Table 1.
*
Signicance at the 10% level.
**
Signicance at the 5% level.
***
Signicance at the 1% level.

regulations are discriminatory in nature, with large elite SOEs

receiving favorable treatment and protection from the state. An
important privilege enjoyed by SOEs is their ease of access to external nance. This allows them not only to obtain loans from the
state banking system, but also to raise external nance through
seasoned equity offerings and rights issues. As a result of having
less nancial constraints, SOEs hold less cash for forthcoming
payables and make better use of receivables as a cash substitute.
Therefore, like related-party trades, easier access to nance
makes state ownership a substitute for nancial deepening. Thus,
we predict that the effect of state ownership on the sensitivity of
cash holdings to trade credit will be less marked for rms in
regions with lower levels of nancial deepening than for those
in regions with higher levels of nancial deepening. We run similar
regression models to those reported in Table 8 by replacing
related-party trades with state ownership. The results are reported
in Table 9.
As shown in Model (1) of Table 9, the coefcients of the interaction terms between CRDT_PAY and STATE and between CRDT_REV
and STATE in the subsample with lower levels of nancial deepening are both signicantly negative. This suggests that fewer nancing constraints help SOEs to reduce the cash held to cover payables
and to increase the substitute ratio of receivables for cash. However, in Model (2) for the subsample with higher levels of nancial
deepening, the coefcient of the interaction term between

CRDT_PAY and STATE is positive but not signicant, whereas that

of the interaction term between CRDT_PAY and STATE is marginally
signicantly negative. This indicates that in regions with high
levels of nancial deepening, state ownership does not exert a signicant impact on the sensitivity of cash holdings to trade credit.
These results suggest that the effect of state ownership on the sensitivity of cash holdings to receivables is greater among rms in regions with lower levels of nancial deepening than among those in
regions with higher levels of nancial deepening. State ownership
serves as a substitute for nancial deepening in the relation between trade credit and cash holdings.
Model (3) of Table 9 reports the results of models with threeway interaction terms among CRDT_PAY, STATE, and DEEPEN, and
among CRDT_REV, STATE, and DEEPEN. The coefcients of these
three-way interaction terms are both signicantly positive. This
suggests that the differences in the impact of state ownership on
the sensitivity of cash holdings to trade credit between rms in
regions with different levels of nancial deepening are statistically
signicant. The effects are more pronounced in regions with lower
levels of nancial deepening than in those with higher levels of
nancial deepening. These results also support the ndings of
Models (1) and (2) that the roles of state ownership and nancial
deepening are substitutable.
Model (4) of Table 9 presents the results of the interactive effect
of state ownership and the new receivables pledge policy on the

2880

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

Table 8
Related-party trades and the relationship between trade credit and cash holdings.

LIQUID2
SIZE
LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN
CRDT_PAY
CRDT_REV
CRDT_PAY  DRPT_BUY
CRDT_REV  DRPT_SELL

Model 1: Subsample with a lower level of nancial

deepening

Model 2: Subsample with a higher level of nancial

deepening

Model 3

Model 4

0.292***
(0.000)
0.038***
(0.000)
0.419***
(0.000)
0.161***
(0.000)
0.007*
(0.055)
0.416***
(0.000)
0.459***
(0.000)
0.055***
(0.000)
0.006
(0.671)
0.008*
(0.084)
0.040
(0.148)
0.766***
(0.000)
0.098***
(0.000)
0.088***
(0.001)
0.089***
(0.000)

0.430***
(0.000)
0.052***
(0.000)
0.459***
(0.000)
0.216***
(0.000)
0.007*
(0.089)
0.569***
(0.000)
0.535***
(0.000)
0.055***
(0.000)
0.037**
(0.020)
0.006
(0.260)
0.030*
(0.098)
0.785***
(0.000)
0.199***
(0.000)
0.030
(0.365)
0.018
(0.445)

0.359***
(0.000)
0.043***
(0.000)
0.448***
(0.000)
0.192***
(0.000)
0.010***
(0.000)
0.493***
(0.000)
0.483***
(0.000)
0.056***
(0.000)
0.011
(0.299)
0.008**
(0.029)
0.037***
(0.002)
0.799***
(0.000)
0.149***
(0.000)
0.127***
(0.005)
0.103***
(0.001)
0.024**
(0.038)
0.054***
(0.003)

0.359***
(0.000)
0.043***
(0.000)
0.447***
(0.000)
0.192***
(0.000)
0.010***
(0.000)
0.494***
(0.000)
0.483***
(0.000)
0.056***
(0.000)
0.010
(0.331)
0.008**
(0.028)
0.036***
(0.002)
0.800***
(0.000)
0.149***
(0.000)
0.096***
(0.000)
0.055***
(0.001)

CRDT_PAY  DRPT_BUY  DEEPEN

CRDT_REV  DRPT_SELL  DEEPEN
CRDT_PAY  DRPT_BUY  POLICY

0.023
(0.332)
0.051**
(0.025)

CRDT_REV  DRPT_SELL  POLICY

Sample size
Adj-R2

6668
0.356

6561
0.387

13,229
0.367

13,229
0.367

This table reports the results of the regression of related-party trades on the relationship between trade credit and cash holdings. The constant term, industry dummies, year,
and province dummies are included in the regression but are not reported. The p-values, adjusted for clustering at the rm level, are presented in parentheses below the
estimates, where the denitions of the variables are as presented in Table 1.
*
Signicance at the 10% level.
**
Signicance at the 5% level.
***
Signicance at the 1% level.

relation between trade credit and cash holdings. The coefcient of

the three-way interaction term among CRDT_PAY, STATE, and POLICY is not signicant, while that of the interaction term among
CRDT_REV, STATE, and POLICY is signicantly negative. This suggests that the effect of the new receivables pledge policy on the
sensitivity of cash holdings to trade receivables is greater for SOEs
than for non-SOEs, whereas the effect of the new receivables
pledge policy on the sensitivity of cash holdings to trade payables
is no different between SOEs and non-SOEs. Similar to the results
for related-party trades, these results also support our second
hypothesis that the new receivables pledge policy helps rms to
increase the substitute ratio of receivables for cash, but does not
affect the sensitivity of cash holdings to trade payables.
5.7. Two-stage analysis
We conduct ve two-stage instrumental variable regressions
similar to the models presented in Table 5. For each model, we report only the coefcient of the lagged cash holdings estimated in
the rst stage to account for the presence of reverse causality,
and the coefcients of trade payables and receivables (CRDT_PAY

and CRDT_REV) or their interaction term with the nancial deepening measure, which reect the coefcients corrected for endogeneity. The results are reported in Table 10.
The coefcients of lagged cash are signicant in all ve models
in which the dependent variable is trade payables (CRDT_PAY). This
suggests that cash has a positive impact on trade payables, in that
rms that hold more cash are more likely to use trade payables to a
greater extent in their operations. All of the models other than
Model (1) have signicantly negative coefcients on the lagged
cash ratio for the regressions of trade receivables. This provides
some evidence that rms that hold more cash tend to hold less
trade receivables. In common with the results reported in Table
10, the coefcients of CRDT_PAY and CRDT_REV and their secondstage interaction terms are all signicant. Overall, these results
show that our conclusions hold after correcting for the endogeneity problem.
6. Conclusion
In this study, we investigate the relationship between trade
credit and cash holdings and how it is affected by nancial

2881

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

Table 9
State ownership and the relationship between trade credit and cash holdings.

LIQUID2
SIZE
LEV
DEBTM
M/B
CAPEX
CASHFLOW
DIVIDEND
TOP1
STATE
DEEPEN
CRDT_PAY
CRDT_REV
CRDT_PAY  STATE
CRDT_REV  STATE

Model 1: Subsample with a lower level of nancial

deepening

Model 2: Subsample with a higher level of nancial

deepening

Model 3

Model 4

0.274***
(0.000)
0.043***
(0.000)
0.390***
(0.000)
0.145***
(0.000)
0.004
(0.361)
0.348***
(0.000)
0.474***
(0.000)
0.051***
(0.000)
0.014
(0.322)
0.007
(0.407)
0.017
(0.605)
0.642***
(0.000)
0.032
(0.259)
0.069**
(0.028)
0.058**
(0.039)

0.393***
(0.000)
0.051***
(0.000)
0.424***
(0.000)
0.199***
(0.000)
0.001
(0.757)
0.519***
(0.000)
0.523***
(0.000)
0.050***
(0.000)
0.004
(0.795)
0.001
(0.935)
0.002
(0.913)
0.655***
(0.000)
0.148***
(0.000)
0.002
(0.967)
0.036*
(0.064)

0.334***
(0.000)
0.047***
(0.000)
0.413***
(0.000)
0.173***
(0.000)
0.002
(0.608)
0.424***
(0.000)
0.488***
(0.000)
0.051***
(0.000)
0.004
(0.669)
0.003
(0.604)
0.018
(0.227)
0.646***
(0.000)
0.090***
(0.000)
0.047
(0.161)
0.057*
(0.075)
0.034**
(0.032)
0.052**
(0.025)

0.333***
(0.000)
0.044***
(0.000)
0.419***
(0.000)
0.180***
(0.000)
0.008***
(0.002)
0.448***
(0.000)
0.467***
(0.000)
0.056***
(0.000)
0.008
(0.411)
0.002
(0.810)
0.034***
(0.003)
0.685***
(0.000)
0.125***
(0.000)
0.044*
(0.081)
0.030
(0.228)

CRDT_PAY  STATE  DEEPEN

CRDT_REV  STATE  DEEPEN
CRDT_PAY  STATE  POLICY

0.0320
(0.453)
0.134***
(0.000)

CRDT_REV  STATE  POLICY

Sample size
Adj-R2

6668
0.366

6561
0.390

13,229
0.375

13,229
0.366

This table reports the results of the regression of state ownership on the relation between trade credit and cash holdings. The constant term, industry dummies, year, and
province dummies are included in the regression but are not reported. The p-values, adjusted for clustering at the rm level, are presented in parentheses below the
estimates, where the denitions of the variables are as presented in Table 1.
*
Signicance at the 10% level.
**
Signicance at the 5% level.
***
Signicance at the 1% level.

deepening. Using a sample of Chinese listed rms over the period

19992009, we nd that trade payables and receivables exert
an asymmetric effect on cash holdings whereby rms hold an
additional $0.71 of cash for every $1 of credit payable, but use $1
of receivables as a substitute for only $0.15 of cash. This result
is in contrast to the traditional inference that $1 of receivables
can be used to cover $1 of payables, and suggests that trade credit
payables and receivables may have independent and different
effects on cash holdings.
We also examine how nancial deepening inuences the
relationship between trade credit and cash holdings. We nd
that rms in regions with higher levels of nancial deepening
tend to hold less cash for payables, whereas the substitute
ratio of receivables for cash is higher for rms in regions with
greater nancial depth. In addition, we investigate how the
new receivables pledge policy implemented in 2007, which ofcially allows rms to use trade receivables to secure loans,
affects the sensitivity of cash holdings to trade credit. We nd
that the substitute ratio of receivables for cash has increased
signicantly since the receivables pledge policy was implemented.

The impact of the new receivables pledge policy is more

pronounced for rms in regions with higher levels of nancial
deepening. These results suggest that deeper nancial development not only helps rms to obtain nance, but also reduces
the cost of cash shortages for payables and the cost of converting
receivables into cash.
This study adopts a micro-level perspective by linking nancial
deepening with two important corporate operating activities:
trade credit and cash holdings. Our investigation not only complements previous studies on trade credit and cash holdings, but also
extends the literature on nancial deepening. Our results show
that trade credit increases the explanatory power of cash holdings.
The asymmetric inuence of payables and receivables suggests
that they should be considered separately when studying their
impact on cash holdings. In terms of trade credit research, our
ndings deepen the understanding of the inuence of trade
credit on rm operations. Our ndings also add to the literature
on nancial deepening by providing evidence of a micro-level
channel through which nancial deepening affects economic activity and by demonstrating that the development of the nancial

2882

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

Table 10
Two-stage instrumental variable regressions.

First stage
Lagged cash ratio (for
CRDT_PAY)

Model 1: subsample with a lower level of

nancial deepening

Model 2: subsample with a higher level of

nancial deepening

Model 3

Model 4

Model 5

0.081

0.032

0.049

0.049

0.049

***

Lagged cash ratio (for

CRDT_REV)
Second stage
CRDT_PAY
CRDT_REV

***

***

(0.000)
0.008

(0.000)
0.055

(0.000)
0.038

(0.000)
0.038

(0.000)***
0.038

(0.345)

(0.000)***

(0.000)***

(0.000)***

(0.000)***

6.415***
(0.000)
0.738***
(0.000)

4.100***
(0.000)
2.410***
(0.000)

5.301***
(0.000)
1.727***
(0.000)
0.227***
(0.000)

5.478***
(0.000)
1.771***
(0.000)

5.249***
(0.000)
1.577***
(0.000)
0.278***
(0.000)
0.117***
(0.000)

CRDT_PAY  DEEPEN

***

0.045***
(0.001)

CRDT_REV  DEEPEN

This table reports the results of the two-stage instrumental variable regressions. The rst stage of the procedure involves an OLS analysis in which trade payables (or
receivables) are regressed against the same controls used for the OLS regressions and the panel analyses, plus four other variables known to affect trade credit. These variables
are xed assets to assets, the natural logarithm of rm age in years, the lagged cash holding ratio, and a dummy that indicates whether the rm has engaged in related-party
trades in the year. The estimated trade payables (receivables) generated in the rst stage are then included in the second-stage regression in which the dependent variable is
the cash holding measure. For each model, we report only the coefcient of the lagged cash holding measure in the rst stage to check for the presence of reverse causality,
and the coefcient of trade payables and receivables (CRDT_PAY and CRDT_REV) or its interaction term with the nancial deepening measure, which reects the coefcients
corrected for endogeneity. The p-values adjusted for clustering at the rm level are presented in parentheses below the estimates, where the denitions of the variables are as
presented in Table 1.

Signicance at the 10% level.

Signicance at the 5% level.

***
Signicance at the 1% level.

environment can shape the relationship between trade credit and

cash holdings.
This study also has policy implications. Our results suggest that
nancial deepening not only helps rms to obtain nance, but also
improves the function of trade credit as a short-term nancing
instrument. The Chinese government should continue in its effort
to deepen the nancial environment to ease rms nancing constraints and thus promote economic efciency. Our results also
show that the new receivables pledge policy implemented in China
in 2007 complements nancial deepening because it helps rms to
increase the substitute ratio of credit receivables for cash. This efcient new policy indicates there are many things the government
can do to mitigate rms nancial constraints. These include promoting nancial institutions and nancial markets, developing

other nancing channels such as trade credit, and enhancing asset

liquidity.
Acknowledgments
We would like to thank Ike Mathur (the editor), an anonymous
referee, Chendi Zhang and participants at the 2010 Symposium on
International Corporate Finance and Governance held in Twente,
the Netherlands for their helpful comments. Wenfeng Wu
acknowledges nancial support from the National Science Fund
Committee of China (No. 70672074) and the Shanghai Pujiang Program. He also thanks the Robert H. Smith School of Business of the
University of Maryland, College Park for research support during
his visit.

Appendix A
Average nancial deepening ratio by province

PROVINCE

DEEPEN

PROVINCE

DEEPEN

PROVINCE

DEEPEN

PROVINCE

DEEPEN

Anhui
Beijing
Chongqing
Fujian
Gansu
Guangdong
Guangxi
Guizhou

0.757
2.018
1.058
0.737
0.966
0.900
0.723
1.101

Hainan
Hebei
Heilongjiang
Henan
Hubei
Hunan
Jiangsu
Jiangxi

1.065
0.630
0.741
0.678
0.825
0.703
0.773
0.747

Jilin
Liaoning
Neimenggu
Ningxia
Qinghai
Shaanxi
Shandong
Shanghai

0.997
0.970
0.672
1.304
1.208
0.990
0.645
1.404

Shanxi
Sichuan
Tianjin
Xinjiang
Xizang
Yunnan
Zhejiang

0.965
0.887
1.145
0.862
0.667
1.055
1.122

The table lists the average nancial deepening ratio by province from 1998 to 2008. The nancial deepening ratio is calculated as bank
loans/GDP ratio of the province for the year.

W. Wu et al. / Journal of Banking & Finance 36 (2012) 28682883

References
Bates, T., Kahle, K., Rene, S., 2009. Why do US rms hold so much more cash than
they used to? Journal of Finance 54, 19952021.
Bartholdy, J., Mateus, C., 2008. Taxes and Corporate Debt Policy: Evidence for
Unlisted Firms of Sixteen European Countries. Working Paper, <http://
www.ssrn.com/abstract=1098370> (accessed on 05.04.11).
Beck, T., Demirg-Kunt, A., Maksimovic, V., 2004. Financial and legal constraints to
rm growth: does size matter? Journal of Finance 60 (1), 137177.
China Banking Regulatory Commission, 2010. 2009 Annual Report of China Banking
Regulatory Commission. <http://www.cbrc.gov.cn> (accessed on 05.04.11).
China Securities Regulatory Commission, 2011. Monthly Report of Securities
Markets in December 2010. <http://www.csrc.gov.cn> (accessed on 05.04.11).
Cull, R., Xu, C.L., Zhu, T., 2009. Formal nance and trade credit during Chinas
transition. Journal of Financial Intermediation 18, 173192.
Cunat, V., 2007. Trade credit: suppliers as debt collectors and insurance providers.
Review of Financial Studies 20, 491527.
Demirg-Kunt, A., Maksimovic, V., 1998. Law, nance and rm growth. Journal of
Finance 53, 21072131.
Dittmar, A., Mahrt-Smith, J., 2007. Corporate governance and the value of cash
holdings. Journal of Financial Economics 83 (3), 599634.
Dittmar, A., Mahrt-Smith, J., Servaes, H., 2003. International corporate governance
and corporate cash holdings. Journal of Financial and Quantitative Analysis 38,
111133.
Dornbusch, R., Reynoso, A., 1989. Financial factors in economic development.
American Economic Review 79 (2), 204209.
FIAS/IFC, 2007. Reforming Collateral Laws and Registries: International Best
Practices and the Case of China, March, FIAS/IFC PEP Report.
Fishman, R., Love, I., 2003. Trade credit, nancial intermediary development, and
industry growth. Journal of Finance 58, 353374.
Foley, C., Hartzell, J., Titman, S., Twite, G., 2007. Why do rms hold so much cash? A
tax-based explanation. Journal of Financial Economics 86, 579607.
Ge, Y., Qiu, J., 2007. Financial development, bank discrimination and trade credit.
Journal of Banking and Finance 31, 513530.

2883

Giannetti, M., Burkart, M., Ellingsen, T., 2011. What you sell is what you lend?
Explaining trade credit contracts. Review of Financial Studies. 24, 12611298.
Guney, Y., Ozkan, A., Ozkan, N., 2007. International evidence on the non-linear
impact of leverage on corporate cash holdings. Journal of Multinational
Financial Management 17 (1), 4560.
Harford, J., Mansi, S., Maxwell, W., 2008. Corporate governance and rm cash
holdings in the US. Journal of Financial Economics 87, 535555.
Hasan, I., Wachtel, P., Zhou, M., 2009. Institutional development, nancial
deepening and economic growth: evidence from China. Journal of Banking
and Finance 33, 157170.
Jin, H., Qian, Y., Weingast, B., 2005. Regional decentralization and scal incentives:
federalism Chinese style. Journal of Public Economics 89 (910), 17191742.
Li, K., Prabhala, N.R., 2005. Self-selection Models in Corporate Finance. Working
Paper, University of Maryland. <http://ssrn.com/abstract=843105< (accessed on
05.04.11).
Morris, J., 1992. Factors Affecting the Maturity Structure of Corporate Debt. Working
paper, University of Colorado at Denver.
Opler, T., Pinkowitz, L., Stultz, R., Williamson, R., 1999. The determinants and
implications of corporate cash holdings. Journal of Financial Economics 52, 3
46.
Ozkan, A., Ozkan, N., 2004. Corporate cash holdings: an empirical investigation of
UK companies. Journal of Banking and Finance 28 (9), 21032134.
Peoples Bank of China, 2009. Good Performance of the Accounts Receivable Pledge
Registration and Announcement System during the Past Two Years. (9.12.09).
Peoples Bank of China, 2010. Sources & Uses of Credit Funds of Financial
Institutions in 2009. <http://www.pbc.gov.cn> (accessed on 05.04.11).
Petersen, M., 2009. Estimating standard errors in nance panel data sets: comparing
approaches. Review of Financial Studies 22, 435480.
Petersen, M., Rajan, R., 1997. Trade credit: theories and evidence. Review of
Financial Studies 10, 661697.
Rajan, R., Zingales, L., 1995. What do we know about capital structure? Some
evidence from international data. Journal of Finance 50, 14211460.
Wu, W., Wu, C., Rui, O.M., forthcoming. Ownership and the value of political
connections: evidence from China. European Financial Management.