Professional Documents
Culture Documents
Price College of Business, The University of Oklahoma, 307 West Brooks, 205A Adams Hall, Norman, OK 73019-4005, United States
King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia
c
College of Business Administration, University of Texas at El Paso, El Paso, TX 79968, United States
b
a r t i c l e
i n f o
Article history:
Received 22 June 2013
Accepted 13 June 2014
Available online 8 July 2014
JEL classication:
G32
G34
G38
Keywords:
Privatization
State ownership
Soft-budget constraints
Cash holdings
China
a b s t r a c t
We study the relation between state ownership and cash holdings in Chinas share-issue privatized rms from
2000 to 2012. We nd that the level of cash holdings increases as state ownership declines. For the average
rm in our sample, a 10 percentage-point decline in state ownership leads to an increase of about RMB 55 million in cash holdings. This negative relation can be attributable to the soft-budget constraint (SBC) inherent in
state ownership. The Chinese nancial system is dominated by the state-owned banks, an environment very
conducive for the SBC effect. We further examine and quantify the effect of state ownership on the value of
cash and nd that the marginal value of cash increases as state ownership declines. The next RMB added to
cash reserves of the average rm is valued at RMB 0.96 by the market. The marginal value of cash in rms with
zero state ownership is RMB 0.36 higher than in rms with majority state ownership. The SBC effect exacerbates
agency problems inherent in state-controlled enterprises, contributing to their lower value of cash.
2014 Elsevier B.V. All rights reserved.
1. Introduction
The efcient management of liquidity is essential to the success
of any business. Over the past decade or so, nancial economists
have devoted much attention to the causes and consequences of
the large and increasing amount of liquid assets on corporate balance sheets around the world. Scholars have developed various
explanations for the cross-section, cross-country, and time-series
variations in corporate cash holdings. These explanations encompass
q
We thank Jarrad Harford, Ivalina Kalcheva, Sifei Li, Karl Lins, Chen Liu, Yixin Liu,
Yu Liu, Neslihan Ozkan, Meijun Qian, Alejandro Riano, Valerity Sibilkov, and Feixue
Xie for helpful comments and suggestions. This study has also beneted from
comments offered by participants in the 2013 ECCE-USB Financial Globalization
and Sustainable Finance conference at the Stellenbosch Business School, Cape
Town, South Africa and the 2012 Financial Management Association meeting in
Atlanta, Georgia. Finally, we appreciate the nancial support afforded to this project
by the University of Oklahomas Price College of Business and George Lynn Cross
Research Professorship, the KFUPM Chair Professor in Finance, and the University of
Texas at El Pasos College of Business Administration.
Corresponding author at: Price College of Business, The University of Oklahoma,
307 West Brooks, 205A Adams Hall, Norman, OK 73019-4005, United States. Tel.: +1
(405) 325 2058; fax: +1 (405) 325 7688.
E-mail addresses: wmegginson@ou.edu (W.L. Megginson), sbullah@utep.edu
(B. Ullah), zwei@utep.edu (Z. Wei).
1
Tel.: +1 (915) 400 8478.
2
Tel.: +1 (915) 747 5381.
http://dx.doi.org/10.1016/j.jbankn.2014.06.011
0378-4266/ 2014 Elsevier B.V. All rights reserved.
277
We choose to study the SBC effect, as proxied by state ownership, on corporate cash holdings in Chinas SIP rms for several reasons. First, there is a wide cross-sectional variation of state
ownership in Chinas SIP rms, ranging from zero to over 90% in
our sample rms. This wide variation allows us to examine the
degree of the SBC effect on the level of cash holdings, as well as
on the marginal value of cash. Second, cash holdings in Chinas
publicly traded rms are large and increasing. At the end of
2012, the total amount of liquid assets held by Chinas publicly
traded non-nancial rms in our sample was approximately RMB
2.95 trillion (about US$ 469 billion).6 Fig. 1 illustrates the mean cash
holdings and state ownership over our sample period 20002012.
Fig. 1 shows dramatic decreases in state ownership over the sample
period, from a mean of 34.7% in 2000 to 4.3% in 2012. In contrast, the
cash ratio increases signicantly over the same time period, from a
mean of 18.7% in 2000 to 32.8% in 2012. 7 Fig. 1 illustrates an overall
negative relation between state ownership and cash holdings. Third,
as pointed out earlier, Chinas SIP rms operate in a nancial system
dominated by state-owned banks. Such an environment is very conducive to the SBC effect (Lin and Tan, 1999). However, scholars thus
far have paid little attention to the causes and consequences of cash
holdings in Chinas listed sector, even though China has the worlds
second largest economy and one of the largest stock markets (Allen
et al., 2005). In this regard, our study contributes to the cash holding
literature by providing the rst comprehensive empirical evidence
from China.
We rst examine the relation between state ownership and
cash holdings. We nd that state ownership and cash holdings
are negatively related, consistent with our hypothesis that high
state ownership leads to SBC, and therefore less need to hold high
levels of cash. For the average rm in our sample, a 10 percentagepoint decline in state ownership leads to an increase of about RMB
55 million in cash holdings. This nding is robust to a battery of
robustness checks, including year effects, rm xed effects, and
controlling for Chinas split share structure reform and the potential endogeneity between state ownership and cash holdings. We
further nd that smaller, more protable, and high growth rms
hold more cash, and that debt and net working capital are negatively related to cash holdings, consistent with ndings in U.S.
and other international rms.
In the second part of this paper, we examine the effect of state
ownership on the marginal value of cash, and nd that the
marginal value of cash increases as state ownership declines. For
the average rm in our sample, the next RMB of cash added to its
cash reserves is valued at RMB 0.96, whereas the next RMB of cash
is valued RMB 0.36 higher in rms with zero state ownership than
in rms with majority state ownership. As the government reduces
its stake in the SIP rms, access to credit in state-owned banks
becomes harder, so the budget constraint hardens. Firms with hard
budget constraints invest their funds more efciently and receive
higher returns, leading to higher marginal values of cash (Kornai,
2001; Denis and Sibilkov, 2010).
The rest of the paper is organized as follows. Section 2 provides
a brief discussion of the SBC theory and a snapshot of the Chinese
nancial system. Section 3 describes our data and sample, while
Section 4 studies the relation between state ownership and the
level of cash holdings. Section 5 examines the effect of state ownership on the marginal value of cash, while Section 6 presents
conclusions.
6
Liquid assets (or cash reserves) include cash and marketable securities at the end
of 2012 across all non-nancial rms in our database. The Chinese currency is called
renminbi (peoples money), or RMB. The exchange rate at the end of 2012 was RMB
6.2896/US$ (IMF International Financial Statistics).
7
The cash ratio is dened as cash reserves over total book assets minus cash
reserves (i.e. noncash assets).
278
0.40
0.35
0.30
0.25
0.20
0.15
Mean cash holdings
0.10
Mean state ownership
0.05
0.00
1999
2001
2003
2005
2007
2009
2011
2013
Year
Fig. 1. State ownership and cash holdings in China, 20002012. Cash holdings is cash ratio, measured as cash and cash equivalents divided by net assets. Net assets is dened
as total book assets minus cash and cash equivalents. State ownership is the fraction of total shares owned by the state and managed by Chinas State Assets Management
Authority (SAMA).
8
Please see Kornai et al. (2003) for a detailed review of the SBC theory and related
empirical evidence.
279
Listed rms
N
Market cap
(RMB billion)
Corp bonds
(RMB billion)
Bank loans
(RMB billion)
NPL
%
Exchange rate
RMB/US$
Ination (%)
CPI change
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
1086
1154
1223
1285
1373
1378
1434
1550
1635
1718
2063
2342
2494
4809
4352
3833
4246
3706
3243
8940
32,714
12,137
24,394
26,542
21,476
23,036
86
101
133
169
202
402
553
768
1325
2554
3632
4930
7463
9937
11,231
13,129
15,900
17,820
19,469
22,535
26,169
30,339
39,968
47,920
54,795
62,991
22.4
29.8
26
20.4
15.6
8.6
7.1
6.2
2.4
1.6
1.1
1.0
0.9
8.2783
8.2770
8.2770
8.2770
8.2765
8.0702
7.8087
7.3672
6.8346
6.8282
6.6229
6.3009
6.2896
0.35
0.73
-0.77
1.15
3.88
1.82
1.46
4.75
5.86
-0.7
3.31
5.41
2.62
The table presents descriptive statistics of the major components of the Chinese nancial market from 2000 to 2012: the stock market, the bond market, and the banking
sector. Ination is measured as percentage change in Chinas CPI. NPL is non-performing loans to total loans. RMB/US$ exchange rates are the year-end rates from the IMF.
Policy bonds are bonds issued by policy banks owned by the Treasury Department for specic infrastructure projects.
Data sources: Statistical Yearbook of China 20002013 (Statistical Bureau of China); China Financial Stability Report 20072013 (The Peoples Bank of China). Chinas CPI and the
RMB/US$ exchange rates are from IMF International Financial Statistics.
9
Ayyagari et al. (2010) and Allen et al. (2011) provide excellent reviews of the
institutional details of the Chinese nancial system.
10
Statistical Yearbook of China, 2013, China State Statistical Bureau (Beijing, China).
280
equity while the median is 0.05%. The mean change in net external
funds raised (DNFt) is 1.93% of lagged market equity while the
median is zero. The mean change in cash dividends (DDIVt) is close
to zero and the median is zero. The mean market leverage (MLEVt)
is 15.75% and median is 11.33%. The distributional characteristics
of most of these variables are similar to the U.S. based data, as
found in Faulkender and Wang (2006) and others, although the
magnitude is different for some variables. While the U.S. and
Chinese data are not directly comparable, it is helpful to gain a
sense of what the China data look like.
14
We also divide the sample rms into ve groups according to which geographical
regions their headquarters are located: Northeast, Coastal, Central, Southwest, and
Northwest. The location dummies are included in our pooled OLS regression to
control for regional disparities in economic development. The location classications
and summary statistics are available in an online appendix accompanying this article.
281
Denition
Mean
25th percentile
50th percentile
75th percentile
Stdev
16,441
16,441
16,441
16,441
16,441
16,441
16,436
16,436
15,884
16,441
16,441
16,441
16,441
16,441
16,441
16,441
16,441
16,441
0.2432
0.2139
6029.0
6569.5
0.4216
0.5826
0.3382
0.2433
0.2887
-0.0296
0.0587
0.0736
0.3154
7.7841
0.56
0.1602
0.0184
0.1022
0.0882
0
927.4
1485.7
0.2202
0.4236
0.1986
0.1009
0.1101
-0.1859
0.0032
0.0192
0.0166
4
0
0
0
0.0855
0.1587
0.0569
1742.8
2592.6
0.3641
0.5862
0.2977
0.2357
0.2555
-0.0264
0.0521
0.0521
0.1510
7
1
0.0210
0
0.0967
0.2908
0.4354
3705.3
4988.0
0.5634
0.7346
0.4375
0.3641
0.4207
0.1367
0.1096
0.1052
0.3105
11
1
0.2900
0
0.1184
0.2622
0.2497
41215.3
43132.9
0.2662
0.2250
0.1895
0.1715
0.2257
0.2465
0.1057
0.0720
0.9180
4.7289
0.5
0.2210
0.0755
0.0244
r i;t RBi;t
16,444
0.0505
0.2790
0.0955
0.0984
0.5707
STATE
DCASHt
CASHt1
DEBITt
DNAt
DINTt
DDIVt
DNFt
MLEVt
16,444
16,444
16,444
16,444
16,444
16,444
16,444
16,321
16,439
0.2140
0.0015
0.1485
0.0000
0.0662
0.0013
0.0007
0.0193
0.1575
0
0.0436
0.0484
0.0196
0.1597
0.0024
0.0016
0
0.0377
0.0569
0.0046
0.0985
0.0011
0.0758
0.0005
0
0
0.1133
0.4354
0.0556
0.1855
0.0211
0.3178
0.0049
0.0044
0.0034
0.2411
0.2497
0.1522
0.1609
0.0571
0.6258
0.0129
0.0157
0.0584
0.1485
r i;t RBi;t is excess return, where r i;t is rm is monthly-compounded annual stock return in year t (scal year-end) and RBi;t is stock is benchmark portfolio return at year t. All
variables except excess return and MLEV are deated by lagged market value of equity, MVEt1. DXt = Xt Xt1. CASH is cash plus cash equivalents. EBIT is operating income
before interest and taxes. NA is book assets less cash and cash equivalents. INT is interest expenses. DIV is cash common dividends. NF is net external funds raised, equal to
the sum of net proceeds from equity and debt issuance. MLEV is market leverage, dened as short-term debts plus long-term debts, scaled by market value of equity.
Subscript (t-1) means the beginning of scal year t.
282
Table 3
Cash holdings and state ownership subsamples statistics.
Year
Cash holdings
(1)
Mean
(2)
Median
State ownership
Mean
Median
Mean
Median
0.187
0.228
0.212
0.202
0.192
0.189
0.184
0.201
0.207
0.259
0.288
0.336
0.328
0.128
0.162
0.154
0.147
0.135
0.134
0.131
0.139
0.147
0.181
0.187
0.198
0.199
0.347
0.356
0.356
0.357
0.348
0.333
0.283
0.247
0.215
0.120
0.087
0.057
0.043
0.384
0.394
0.397
0.396
0.385
0.360
0.293
0.250
0.186
0.000
0.000
0.000
0.000
(3)
Mean
Panel B: Cash holdings and state ownership before, during and after the split-share reform
Cash
5673
0.195
0.139
3533
0.191
State
5673
0.352
0.389
3533
0.287
(4)
Median
0.135
0.298
Difference
(5)
Mean
(6)
Median
(5)(1)
Mean
7975
7975
0.291
0.095
0.184
0
0.095
0.256
eliminate dividends to meet cash shortfalls. If this is the case, dividend-paying rms are expected to hold less cash than non-paying
rms. On the other hand, rms rarely cut dividends, much less
eliminate them, unless they are in deep nancial distress. Dividend-paying rms may hold more cash in anticipation of the dividends to be paid in the coming period.
Lastly, we include institutional and foreign ownership as control variables. Shleifer and Vishny (1986) argue that institutional
shareholders help mitigate free rider problems and reduce managerial opportunism by forcing managers to disgorge excess cash.
Institutional shareholders also act to promote their own interest,
which can differ from that of other shareholders (Shleifer and
Vishny, 1997). Empirically, Harford et al. (2008) nd no signicant
relation between cash holdings and institutional ownership after
controlling for lagged cash holdings. In the case of China, many
of the institutional owners are owned or controlled by various levels of government, and they also have a prot maximizing motive
(Wei et al., 2005). These two characteristics point to a negative
relation between institutional ownership and cash holdings. Several studies have found that foreign ownership is positively related
to cash holdings. DeFond et al. (2011) examine a multi-country
sample of rms and nd a positive relation between institutional
foreign ownership and the level of corporate cash holdings. Luo
and Hachiya (2005) study cash holdings in Japanese rms and nd
that foreign ownership is positively related to cash holdings.
Dahlquist and Robertsson (2001) study foreign ownership and rm
characteristics in Swedish rms and nd that foreign shareholders
prefer rms with large cash positions on their balance sheets.
These researchers argue that a high cash ratio indicates a good
liquidity position, a good level of protability, and low information
asymmetry, all preferred by foreign investors. Therefore, we expect
a positive relation between cash holdings and foreign ownership in
our analysis.
There are two methods commonly used to estimate similar
models in the corporate nance literature. One is the pooled ordinary least squares (Pooled OLS) regression controlling for industry
and year effects, and the other is the panel regression controlling
for rm xed effects (Panel FE). Though we also present the Pooled
OLS results, we base our inferences on the Panel FE results. Our
main regression model (1) includes rm xed effects, ai, which
help eliminate time-invariant omitted variable bias, and year xed
effects, kt, which help control for economy-wide yearly uctuations. The reported t-statistics are based on robust standard errors
adjusted for potential heteroscedasticity.
4.2. Empirical results
The empirical results are presented in Table 4. Column (1) presents the Pooled OLS regression results, while columns (2)(4)
present the Panel FE regression results using different leverage
measures. Columns (1) and (2) employ short-term debts plus
long-term debts scaled by net assets (Debts) as leverage measure,
while columns (3) and (4) use total liabilities scaled by net assets
(TL) and no-debt liabilities scaled by net assets (NDL) as leverage
measures, respectively. We observe several consistent ndings in
Table 4. The main explanatory variable, state ownership, State,
has a negative and highly signicant coefcient in all of the four
models, consistent with our hypothesis that rms with high state
ownership hold less cash. State has a coefcient of 0.119 in column (1) and 0.092 in column (2). These results suggest that, for
the typical rm in our sample with an average real net assets of
RMB 6026 million (Table 3, Panel B), a 10 percentage-point
decrease in state ownership leads to about RMB 71.7 million (or
US$ 11.5 million) increase in cash holdings in the Pooled OLS
model in column (1) and RMB 55.5 million (or US$ 8.7 million) in
the Panel FE model in column (2).15
The results further show that smaller rms and younger rms
hold more cash. Debts and net working capital are signicantly
15
For example, for a 10 percentage-point decrease in state ownership, the change in
cash holdings is (Pooled OLS): (0.119) (0.10) = 0.0119, or 1.19% of real assets, all
other variables are evaluated at means. The average rm in our sample has real assets
of RMB 6029 million. This translates into 0.0119 6029 = RMB 71.7 million increase in
cash holdings, or about US$ 11.5 million in 2012 exchange rate.
283
State
Log (real size)
Log (rm age)
Pooled OLS
Panel FE
(1)
(2)
(3)
(4)
0.1196
(9.67)
0.0449
(11.82)
0.0752
(16.17)
0.0918
(7.00)
0.0371
(5.93)
0.0468
(8.30)
0.0953
(7.42)
0.0825
(11.99)
0.0372
(6.61)
0.3588
(14.90)
0.0810
(6.33)
0.0579
(9.18)
0.0500
(8.93)
0.2049
(9.39)
0.2887
(11.44)
TL
Debts
NDL
CapEx
Dividend
Cash ow
NWC
Growth
INST
Foreign
Industry sigma
Constant
Location dummies
Industry dummies
Year dummies
Firm xed effects
Obs.
R2 overall
0.1124
(2.77)
0.0595
(10.99)
0.5907
(16.66)
0.0862
(4.19)
0.0068
(2.31)
0.1228
(8.09)
0.1260
(2.53)
1.0149
(9.13)
1.2565
(14.32)
Yes
Yes
Yes
No
16,436
0.335
0.1279
(3.94)
0.0305
(7.61)
0.2885
(11.24)
0.2366
(10.74)
0.0088
(3.07)
0.0936
(6.27)
0.1441
(2.50)
0.7865
(8.82)
1.0859
(8.33)
No
No
Yes
Yes
16,436
0.286
0.1957
(5.98)
0.0392
(9.75)
0.3785
(15.51)
0.0253
(1.21)
0.0071
(2.40)
0.1101
(7.51)
0.1700
(3.27)
0.7916
(9.00)
1.7615
(12.65)
No
No
Yes
Yes
16,441
0.208
0.3338
(13.53)
0.1823
(5.64)
0.0348
(8.82)
0.2926
(11.77)
0.0905
(4.92)
0.0063
(2.21)
0.0933
(6.37)
0.1563
(3.00)
0.7124
(8.30)
1.3693
(10.45)
No
No
Yes
Yes
16,436
0.288
The dependent variable is cash holdings (Cash) measured as cash plus cash equivalents over net assets. Net assets is total books assets minus Cash. State is state ownership.
Real size is net assets adjusted using Chinese CPI into 2012 RMB. Firm age is the number of years since its IPO. Debts is the sum of short-term debt and long-term debt scaled by
net assets. CapEx is Capital expenditures scaled by net assets. Dividend is a dummy equal to one if a rm pays cash dividends and zero otherwise. Cash ow is net cash ow
from operating activities scaled by net assets. NWC is current assets less cash and equivalents minus current liabilities, scaled by net assets. Growth is the average sales growth
rate over the previous three-year period. Sigma is the standard deviation of an industrys cash ow over the sample period. INST is the fraction of shares owned by non-SAMA
legal entities. Foreign is foreign ownership. Geo location dummies are indicator variables correspond to ve geographic locations: Northeast, Coastal, Central, Southwest, and
Northwest. t-Statistics (in the parentheses) are based on robust standard errors. *** Signicant at the 1% level. ** Signicant at the 5% level. * Signicant at the 10% level.
284
Table 5
State ownership, bank loans, and cash holdings Panel FE regression results.
(1)
(2)
0.1063
(7.31)
0.1000
(7.81)
0.1645
(4.68)
0.0488
(6.62)
0.1207
(8.15)
0.1065
(5.63)
0.0500
(23.54)
0.0812
(24.85)
0.0946
(2.98)
0.0595
(15.94)
0.6131
(24.11)
0.0431
(3.81)
0.0066
(3.06)
0.0410
(4.16)
0.0981
(3.34)
0.7983
(8.61)
1.3852
(28.23)
15,884
0.334
Log(real size)
Log(rm age)
CapEx
Dividend
Cash ow
NWC
Growth
INST
Foreign
Industry sigma
Constant
Obs.
R2 overall
State
(%)
Change
in bank
loans (%)
Effect
on cash
ratio
(%)
Mean net
assets
(RMB
million)
Change
in cash
(RMB
million)
Change
in cash
(US$
million)
75.0
53.2
31.4
7.3
Change in
cash (US$
million)
1
0
5.0
1.2
+10
+10
6094.7
4611.1
305.9
55.7
48.6
8.8
The dependent variable is cash holdings (Cash) measured as cash plus cash equivalents over net assets. Net assets is total books assets minus Cash. State is state
ownership. Real size is net assets adjusted using Chinese CPI into 2012 RMB. Firm age
is the number of years since its IPO. Bank loans is net proceeds from bank borrowings (Cash Flow Statement item), scaled by net assets. CapEx is Capital
expenditures scaled by net assets. Dividend is a dummy equal to one if a rm pays
cash dividends and zero otherwise. Cash ow is net cash ow from operating
activities scaled by net assets. NWC is current assets less cash and equivalents
minus current liabilities, scaled by net assets. Growth is the average sales growth
rate over the previous three-year period. Sigma is the standard deviation of an
industrys cash ow over the sample period. INST is the fraction of shares owned by
non-SAMA legal entities. Foreign is foreign ownership. Year effects and rm xed
effects are controlled for in both models. t-Statistics (in the parentheses) are based
on robust standard errors. *** Signicant at the 1% level. ** Signicant at the 5%
level. * Signicant at the 10% level.
The average net assets in rms with and without state ownership are RMB 6094.7
million and RMB 4611.1 million, respectively. Take Panel B, rst row (at 75% state
ownership), as an example of the calculations of the marginal effect of the interaction of bank loans and state ownership. In a rm with 75% state ownership (high),
a 10 percentage-point increase in bank loans results in (see Panel A, column (1) for
the coefcients): 0.1063 (0.75) 0.1000 (+0.10) + 0.1645 (0.75) (+0.10) =
0.077, or 7.7% of net assets, or 0.077 6094.7 = RMB-471.6 million. All other
independent variables are evaluated at their means.
non-debt liabilities (NDL), all scaled by net assets. We perform separate panel FE regressions using TL and NDL as alternative measures of leverage, as shown in columns (3) and (4).
We observe that the coefcient of total liabilities (TL) in column (3) is positive and signicant, consistent with Chen et al.
(2012). Interestingly, column (4) shows that cash holdings is also
signicantly positively related to non-debt liabilities (NDL). One
possible explanation is that rms with high level of non-debt liabilities (NDL) need to hold high level of cash to meet their precommitted obligations. Though the relation between cash holdings and leverage is not a main focus of the paper, these ndings
are interesting and are new contributions to the cash holdings
literature.
4.3. State ownership, bank loans, and cash holdings
The main proposition in our cash holdings analysis is that state
ownership leads to soft-budget constraint (SBC), and rms with
SBC hold less cash. As discussed earlier, Chinese nancial system,
dominated by state-owned banks, is fertile ground for SBC syndromes. The interaction between state ownership and bank loans
provides cross-section identication for our hypothesis. At high
state ownership, rms have relatively softer budget constraints,
and hence rely more heavily on loans from state-owned banks
for liquidity needs. Conversely, at low state ownership, rms have
relatively harder budget constraints, and hence could not rely on
bank loans for liquidity needs. In sum, a rms cash holdings
should be more sensitive to change in bank loans at high state
ownership and less sensitive at low state ownership. To test this
proposition, we add the interaction term between state ownership
and bank loans, Statei,t Bankloansi,t, to Eq. (1):
285
State
Firm age
Debts/net assets
Capital expenditures/net assets
Dividend dummy
Cash ow/net assets
Net working capital/net assets
Growth
Institutional ownership
Foreign ownership
Industry sigma
Constant
Obs
R2 overall
AR(1) test (p-value)
AR(2) test (p-value)
Hansen test of overidentication (p-value)
diff-GMM
(1)
(2)
(3)
(4)
0.0924***
(6.37)
0.0323**
(2.16)
0.0872***
(10.07)
0.0408***
(6.44)
0.0783***
(11.7)
0.2538***
(8.98)
0.0948***
(2.56)
0.0310***
(6.90)
0.2976***
(10.61)
0.2439***
(10.14)
0.0062**
(2.15)
0.0175
(0.98)
0.0973
(1.44)
0.6197***
(6.12)
1.1676***
(8.84)
12,904
0.3127
0.0561***
(4.08)
0.1104***
(2.88)
0.0533***
(8.51)
0.0532***
(9.63)
0.0841***
(2.94)
0.0409*
(1.74)
0.0604
(0.78)
0.1405***
(2.91)
0.0557
(1.54)
0.1099
(0.49)
0.1433**
(2.09)
0.0089**
(1.95)
0.1361*
(1.88)
1.3714*
(1.87)
1.4194***
(4.91)
Reform
Log (real size)
Reduced form
0.0331***
(5.21)
0.0476***
(7.69)
0.3017***
(10.48)
0.1095***
(2.90)
0.0286***
(6.35)
0.2862***
(10.03)
0.2547***
(10.39)
0.0075***
(2.50)
0.0774***
(4.38)
0.1524**
(2.13)
0.7830***
(7.57)
1.0164***
(7.64)
12,904
0.2854
0.3849***
(15.36)
0.1729***
(8.74)
0.0094***
(3.18)
0.062***
(4.07)
0.1411***
(2.54)
0.6268***
(6.53)
1.3855***
(10.56)
16,441
0.2132
14,883
0.000
0.322
0.503
The dependent variable is Cash measured as cash plus cash equivalents over net assets. Net assets is books assets minus Cash. State is state ownership. Reform is an indicator
variable equal to one if rm-years are from 2000 to 2004, and zero if rm-years are from 2008 to 2012. Real size is net assets adjusted using Chinese CPI into 2012 RMB. Firm
age is the number of years since its IPO. Debts is the sum of short-term debt and long-term debt scaled by net assets. CapEx is capital expenditures scaled by net assets.
Dividend is a dummy equal to one if a rm pays cash dividends and zero otherwise. Cash ow is net cash ow from operating activities scaled by net assets. NWC is current
assets less cash and equivalents minus current liabilities, scaled by net assets. Growth is the average sales growth rate over the previous three years period. Industry sigma is
the standard deviation of an industrys Cash ow over our sample period. INST is institutional ownership. Foreign is foreign ownership. Year effects and rm xed effects are
controlled for in all models. t-Statistics (in the parentheses) are based on robust standard errors.
*
Signicant at the 10% level.
**
Signicant at the 5% level.
***
Signicant at the 1% level.
See Liao et al. (2011) for detailed descriptions of the split-share reform process.
that the post-reform cash ratio is higher than that pre-reform, suggesting this uptrend may be partially due to the reform. As rms sell
more previously non-tradable shares to the public, their cash holdings is likely to increase. To account for the potential effect of the
reform on cash holdings, we exclude the rm-years observations
in 2005, 2006, and 2007, and re-estimate our main model (1) with
panel FE regression. The results from the truncated sample, as shown
in column (1) of Table 6, are similar and consistent with those in
Table 4. In particular, the coefcient of State (0.0924) is almost
identical to the coefcient of State in Table 4 (column (2)) (0.0918).
Chen et al. (2012) study cash holdings surrounding the split share
reform in Chinas listed rms from 2000 to 2008. They document a
signicant decrease in cash holdings following the split share reform.
Their ndings seem to contradict the time series pattern in Fig. 1 of
this study, which shows a dramatic uptrend in cash holdings postreform. Though the main focus of our paper is the relation between
state ownership and cash holdings, the effect of the reform on cash
holdings deserves further examination. As shown in Table 3 (Panel
B), a univariate test nds that mean cash holding post-2007 is significantly higher than the pre-2005 mean. To investigate the effect of
the reform on cash holdings in a multivariate setting, we construct
a dummy variable, Reform, that equals one if an observation falls in
the post-reform years, 20082012, and zero if an observation falls
in the pre-reform years, 20002004. Firm-year observations in
286
DCASHi;t
DEBITi;t
DNAi;t
DINTi;t
b2
b3
b4
MVEi;t1
MVEi;t1
MVEi;t1
MVEi;t1
DNFi;t
DDIVi;t
CASHi;t1
b5
b6
b7
b8 MLEVi;t
MVEi;t1
MVEi;t1
MVEi;t1
CASHi;t1 DCASHi;t
DCASHi;t
b9
b10 MLEVi;t
MVEi;t1
MVEi;t1
MVEi;t1
DCASHi;t
b11 STATEi;t b12 STATEi;t
ei;t
3
MVEi;t1
ri;t RBi;t b0 b1
18
We thank one of the anonymous reviewers for suggesting that we estimate a
static model that does not include the lagged dependent variable as independent
variable. We employ the xtabond2 procedure in STATA as detailed in Roodman
(2009).
19
We also estimate equation (1) using the system GMM estimator (Blundell and
Bonds, 1998), and obtain a negative and signicant coefcient for state ownership.
The results are available in the online appendix accompanying this article.
20
Dittmar and Mahrt-Smith (2007) also employ a modied version of the
Faulkender and Wang (2006) model in their examination of governance and value
of cash.
DCASHi;t
.
MVEi;t1
CASHi;t1
MVEi;t1
DCASHi;t
MVE
i;t1
and
21
We choose to construct 16 benchmark portfolios for the Chinese market instead
of the 25 benchmark portfolios in Fama and French (1993). This is because in the
Chinese stock market, the number of listed rms is much smaller than that in the U.S.
stock market. The steps of constructing our benchmark portfolios are detailed in the
online appendix accompanying this article.
287
288
Table 7
Value of cash regression results: full sample.
(1)
Panel A: Regression results
DCASHt
1.2431***
(18.68)
DEBITt
1.4535***
(16.58)
DNAt
0.0011
(0.10)
DINTt
1.0617**
(2.21)
DDIVt
0.3759
(1.37)
DNFt
0.0280
(0.35)
CASHt1
1.2118***
(25.77)
MLEVt
1.6440***
(38.27)
DCASHt CASHt1
0.3384***
(3.20)
DCASHt MLEVt
1.0128***
(5.81)
STATEt
0.2011***
(8.94)
DCASHt STATEt
0.3610***
(3.15)
Constant
N
R2 overall
Mean value
0.1931***
(17.64)
16,154
0.0976
(1)
Port-ave
(2)
ACW
(3)
1.1779***
(15.99)
0.2004***
(7.24)
0.0201*
(1.77)
0.6745*
(1.89)
2.2326***
(7.35)
0.1503**
(2.45)
0.7451***
(14.42)
1.5747***
(34.13)
0.5270***
(4.51)
0.9551***
(5.00)
0.2609***
(10.52)
0.5161***
(4.06)
1.4562***
(17.05)
0.26174***
(7.93)
0.3852***
(26.90)
2.5335***
(3.94)
2.4983***
(6.67)
0.5397***
(5.09)
1.9398***
(31.53)
2.8526***
(49.48)
0.0572
(0.41)
2.1117***
(9.45)
0.12657***
(4.12)
0.9876***
(6.60)
0.1354***
(11.31)
16,316
0.0438
0.0146
(0.99)
16,314
0.0975
(2) Port-ave
(3) ACW
0.1485
0.1576
0.2140
0.90***
The dependent variable is excess return. All variables except excess return, MLEV
and STATE are deated by lagged market value of equity MVEt1. DXt = Xt Xt1.
CASH is cash plus cash equivalents. EBIT is operating income before interest and
taxes. NA is book assets less cash and cash equivalents. INT is interest expenses. DIV
is cash common dividends. NF is net external funds raised, equal to the sum of net
proceeds from equity and debt issuance. MLEV is market leverage, dened as shortterm debts plus long-term debts, scaled by market value of equity. STATE is state
ownership. Subscript (t-1) means the beginning of scal year t. t-Statistics are based
on robust standard errors and are in the parentheses. Model (1) is our base model
which denes change in cash as the realized change in cash from year end (t-1) to
year end t. Models (2) and (3) use alternative denitions of change in cash. Model
(2) uses the benchmark portfolio average as the normal change in cash holdings
(Port-ave), while model (3) uses the Almeida et al. (2004) (ACW) method to
estimate the normal change in cash holdings. The unexpected change in cash is the
realized change minus the normal change. t-Statistics are based on robust standard
errors and are in the parentheses.
Models (1)(3) are evaluated at the mean values of the explanatory variables.
Change of cash, DCASH, takes the value of 1.00, indicating one RMB is added to a
rms cash reserve.
*
Signicant at the 10% level.
**
Signicant at the 5% level.
***
Signicant at the 1% level.
DCASHi;t
OCFi;t1
b0 b1
b2 Growthi;t1 b3 Sizei;t1
MVEi;t1
MVEi;t1
CAPEXi;t1
DNWCi;t1
DSTDBTi;t1
b4
b5
b6
ei;t
MVEi;t1
MVEi;t1
MVEi;t1
4
where OCF is net cash ow from operating activities, CAPEX is capital expenditures, DNWC is change in net working capital, and
DSTDBT is change in short-term debt, all lagged and scaled by
lagged market equity. Growth is the average sales growth rate over
the previous three years and Size is natural logarithm of book
assets.22 We use a two-step OLS to estimate Eq. (4). The rst step
is to estimate the coefcients in Eq. (4) for each industry, while
the second is to use the coefcients estimated in the rst step to calculate the predicted (expected) change in cash for each individual
rm in that industry. The unexpected change in cash is therefore
the realized change in cash minus the change in cash estimated from
Eq. (4). We proceed to re-estimate Eq. (3) using the ACW measure of
unexpected change in cash holdings, and the regression results are
reported in column (3) of Table 7. These results are largely consistent
with those in column (1), in that the marginal value of cash declines
with higher leverage and higher state ownership. Again, our main
prediction remains robust in that state ownership has a signicant
and negative effect on the marginal value of cash. For the average
rm in our sample, one extra RMB of cash raised is valued at RMB
0.90, as shown in Panel B of Table 7.
5.2.2. State ownership and the value of cash: subsample regression
results
The results from the full sample regressions show that state
ownership has a negative effect on the marginal value of cash, as
shown in Table 7. State ownership ranges from zero to almost
90% in our sample. In the full sample of 16444 rm years, 7410
rm years have zero state ownership, 5824 rm years have between
zero and 50% state ownership, and 3210 rm years have 50% or
higher state ownership. Studies have shown that ownership and rm
value have a nonlinear relation (Demsetz and Lehn, 1985; McConnell
and Servaes, 1990; Wei et al., 2005). The SBC theory predicts that the
relative softness of the budget constraint should manifest itself in
the marginal value of cash across different ranges of state ownership.
To perform this robustness check, we divide the full sample into four
groups. Group 1 consists of 7410 rm years with zero state ownership and Group 4 consists of 3210 rm years with 50% or higher state
ownership.23 The remaining rm years are sorted by state ownership
and then divided about equally into two groups: Group 2 has 2912
rm years with 0.029.3% state ownership and Group 3 has 2912 rm
years with 29.350.0% state ownership. We then re-estimate Eq. (3)
across the four groups of rm years, excluding state ownership
(STATEt) and the interaction term (DCASHt STATEt) as explanatory
variables. These stepwise regressions provide another robustness
check on the SBC prediction concerning the marginal value of cash
22
In the original ACW model, lagged Tobins Q is used instead of sales growth rate,
while DNWC, CAPEX and DSTDBT are scaled by lagged book assets instead of lagged
market equity. Acquisitions are also included in their original model. We do not
include acquisitions in our model due to lack of data.
23
Ayyagari et al. (2010) also use 50% as the cutoff point for state-owned vs nonstate owned rms. Though there is no golden share in China, the government can
often control a privatized rm with less than 50% direct ownership though other
state-controlled institutional owners.
289
DEBITt
DNAt
DINTt
DDIVt
DNFt
CASHt1
MLEVt
DCASHt CASHt1
DCASHt MLEVt
Constant
Obs.
R2 overall
Mean variable value
(1)
0% state
(2)
029.3%
(3)
29.350%
(4)
P50%
1.4810***
(14.5)
1.4186***
(9.99)
0.0805***
(4.46)
1.2340*
(1.61)
1.0851**
(2.20)
0.3433***
(2.61)
1.5203***
(16.88)
1.9828***
(24.42)
0.3757**
(2.22)
1.4702***
(5.28)
0.1372***
(8.24)
7287
0.0862
1.3160***
(6.93)
0.4654***
(3.61)
0.0327
(1.13)
0.4174
(0.32)
0.24348
(0.26)
0.6085***
(2.84)
1.4085***
(9.56)
2.3108***
(18.54)
0.1743
(0.60)
1.6757***
(3.65)
0.2876***
(8.93)
2881
0.0893
0.9366***
(4.75)
0.5928***
(5.66)
0.002
(0.09)
0.6336
(0.52)
2.0828***
(2.92)
0.2258
(1.04)
1.3016***
(8.85)
1.7466***
(14.89)
0.0878
(0.29)
0.5759
(1.23)
0.1771***
(5.82)
2836
0.1046
0.9790***
(6.50)
1.9538***
(10.03)
0.067***
(2.57)
0.0203
(0.02)
0.6485
(1.27)
0.4983***
(2.49)
0.9999***
(10.33)
1.3159***
(14.27)
0.2564
(1.16)
0.5135
(1.33)
0.1066***
(5.38)
3157
0.1199
(1)
(2)
(3)
(4)
Diff=(1) - (4)
0.1502
0.1762
0.1374
0.99***
0.1506
0.1769
0.3960
0.82***
0.1368
0.1699
0.6155
0.86***
0.36***
Models (1)(4) are evaluated at the mean values of the explanatory variables. Change of cash, DCASH, takes the value of 1.00, indicating one RMB is added to a rms cash
reserve.
The dependent variable is excess return. All variables except excess return and MLEV are deated by lagged market value of equity MVEt1. DX = Xt Xt1. CASH is cash plus
cash equivalents. EBIT is operating income before interest and taxes. NA is book assets less cash and cash equivalents. INT is interest expenses. DIV is cash common dividends.
NF is net external funds raised, equal to the sum of net proceeds from equity and debt issuance. MLEV is market leverage. Firm years are divided into four groups according to
state ownership in the rm: group 1 in Column (1) has zero state ownership; group 4 in Column (4) has 50% or higher state ownership; the remaining rm years are equally
divided into the middle two groups in Column (2) and Column (3). t-Statistics are based on robust standard errors and are in the parentheses.
*
Signicant at the 10% level.
**
Signicant at the 5% level.
***
Signicant at the 1% level.
6. Conclusions
This study examines the relation between state ownership and
corporate cash holdings in Chinas share-issue privatized rms
from 2000 to 2012. We nd that cash holdings and state ownership
are negatively related, consistent with the prediction of the softbudget constraint theory rst formulated by Kornai (1979, 1980).
This nding is robust after controlling for year effects, rm xed
effects, Chinas split share structure reform, and potential endogeneity. For the average rm in our sample, a 10 percentage-point
decline in state ownership leads to an increase of about RMB 55
million in cash holdings. The level of cash holdings is also negatively related to institutional ownership, possibly due to the fact
that many of the institutional owners are owned or controlled by
various levels of government. We also nd that smaller, more profitable, and higher growth rms hold more cash and that short-term
debts and net working capital are negatively related to cash holdings, suggesting that short-term debts and working capital may be
treated as cash substitutes.
We further examine the relation between state ownership and
the value of cash and nd that the marginal value of cash increases
as state ownership declines. For the average rm in our sample, the
next RMB raised is valued at RMB 0.96 by shareholders. Furthermore, the marginal value of cash in rms with zero state ownership is RMB 0.36 higher than in rms with majority state
290
References
Alfandari, G., Fan, Q., Freinkman, L., 1996. Government nancial transfers to
industrial enterprises and restructuring. In: Commander, Simon, Fan, Qimiao,
Schaffer, Mark (Eds.), Enterprise Restructuring and Economic Policy in Russia.
Economic Development Institute (EDI) Develop. Stud. World Bank, Washington,
DC, pp. 166202.
Allen, F., Qian, J., Zhang, C., Zhao, M., 2011. Chinas nancial system: opportunities
and challenges. In: Fan, J. Morck, R. (Eds.), Capitalizing China. University of
Chicago Press, (forthcoming).
Allen, F., Qian, J., Qian, M., 2005. Law, nance, and economic growth in China.
Journal of Financial Economics 77, 57116.
Almeida, H., Campello, M., Weisbach, M., 2004. The cash ow sensitivity of cash.
Journal of Finance 59, 17771804.
Anderson, J., Korsun, G., Murrell, P., 2000. Which enterprises (believe they) have soft
budgets? Evidence on the effects of ownership and decentralization in
Mongolia. Journal of Comparative Economics 28 (2), 219246.
Arellano, M., Bond, S., 1991. Some tests of specication for panel data: Monte Carlo
evidence and an application to employment equations. The Review of Economic
Studies 58 (2), 277297.
Ayyagari, M., Demirg-Kunt, A., Maksimovic, V., 2010. Formal versus informal
nance: Evidence from China. Review of Financial Studies 23, 30483097.
Bates, T., Kahle, K., Stulz, R., 2009. Why do US rms hold so much cash than they
used to? Journal of Finance 64, 19852021.
Berglf, E., Roland, G., 1998. Soft budget constraints and banking in transition
economies. Journal of Comparative Economics 26 (1), 1840.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic
panel data models. Journal of Econometrics 87 (1), 115143.
Bortolotti, B., W. Megginson, 2012. Privatization trends and major deals of 2011 and
1H2012. Privatization Barometer Report 2011 (Privatization Barometer, Milan).
Boycko, M., Shleifer, A., Vishny, R.W., 1996. A theory of privatization. The Economic
Journal 106, 309319.
Brandt, L., Li, H., 2003. Bank discrimination in transition economies: ideology,
information or incentives. Journal of Comparative Economics 31, 387413.
Chen, Q., Chen, X., Schipper, K., Xu, Y., Xue, J., 2012. The sensitivity of corporate cash
holdings to corporate governance. The Review of Financial Studies 25 (12),
36103644.
Claessens, S., Djankov. S., 1998. Politicians and rms in seven Central and Eastern
European countries. Policy Research working paper #1954, World Bank.
Coricelli, F., Djankov, S., 2001. Hardened budgets and enterprise restructuring:
theory and an application to Romania. Journal of Comparative Economics 29 (4),
749763.
Cull, R., Xu, L., 2000. Bureaucrats, state banks, and the efciency of credit allocation:
the experience of Chinese state-owned enterprises. Journal of Comparative
Economics 28, 131.
Cull, R., Xu, L., 2003. Who gets credit? The behavior of bureaucrats and state banks
in allocating credit to Chinese state-owned enterprises. Journal of Development
Economics 71, 533559.
291
Shleifer, A., Vishny, R., 1994. Politicians and rms. Quarterly Journal of Economics
109 (4), 9951025.
Shleifer, A., Vishny, R., 1986. Large shareholders and corporate control. Journal of
Political Economy 94, 461488.
Shleifer, A., Vishny, R., 1997. A survey of corporate governance. Journal of Finance
52, 737783.
Sun, Q., Tong, W., 2003. China share issue privatization: the extent of its success.
Journal of Financial Economics 70, 183222.
Wang, F., Xu, Y., 2004. What determines Chinese stock returns? Financial Analysts
Journal 60 (6), 6577.
Wei, Z., Xie, F., Zhang, S., 2005. Ownership structure and rm value in Chinas
privatized rms: 19912001. Journal of Financial and Quantitative Analysis 40,
87108.