Pyramidal Ownership Structure, Capital Investment and Firm
Performance：Evidence from China
Chao Chena, Donglin Xiab, Song Zhuc
a
b
c
School of Management, Fudan University
School of Economics and Management, Tsinghua University
School of Economics and Business Administration, Beijing Normal University
Abstract: Business groups organized by pyramids enable the ultimate shareholders to
control a portfolio of firms with less cash requirement. Further, corporate pyramid
induces an internal capital market and makes capital transfer more convenient within
the pyramid. In China, the state and business groups control a large number of listed
firms through pyramidal ownership structure. What role does the corporate pyramid
play in firms’ investment decisions? What is its influence on firm performance? This
paper investigates the capital investment and firm performance from the perspective
of pyramidal ownership structure. We find that as the layers of corporate pyramid
increases, the capital overinvestment declines. The negative relations between
pyramid and overinvestment exist for both state-owned enterprises (SOEs) and
non-state-owned enterprises (NSOEs), which indicate that increasing the layers within
corporate pyramid reduces the likelihood of overinvestment of the listing firm and
improving investment efficiency. Moreover, we show that the effects of increasing the
layers of corporate pyramid on accounting performance are different for SOEs and
NSOES. For SOEs, increasing the layers of pyramid results in less government
interference on the listed firm and more flexibility in operate. Therefore, increasing
pyramidal layers is positively related to accounting performance. While for NSOEs,
pyramiding is to build an internal capital market for the ultimate shareholder’s capital
investment. Although pyramid may reduce overinvestment of the listing firm, agency
costs may offset the positive effect and induce a lower accounting performance.
JEL Classification: G31, G32
Keywords: Pyramidal ownership structure; Capital investment; Internal capital
market; Firm performance
a
Corresponding author, Fudan University, 670 Guoshun Road, Shanghai, China
200433. Phone: 86-21-25011111, Email: [email protected].
1
Pyramidal Ownership Structure, Capital Investment and Firm
Performance：Evidence from China
1. Introduction
In a free market, the optimization of resource allocation is critical for value
creation of a firm. The efficiency of an external capital market can be realized by
allocating capital to the most profitable investment projects (Sudip et al., 2008).
However, several studies on market imperfection and information asymmetry suggest
that the external capital market contents “lemon” premium (Myers and Majluf, 1984),
therefore external financing is much costly than that of internal capital financing due
to market fictions (Brennan and Subrahmanyam, 1996; Easley and O’Hara, 2004).
When facing costly external financing, firms can rely on the internal capital market,
which may reduce the restriction of the fictions on capital. Business groups organized
by corporate pyramids enable the ultimate shareholders to control a portfolio of firms
in the pyramid with limited cash requirement. Thus, it induces an internal capital
market and makes capital transfer convenient within the pyramid. The internal capital
market alleviates restrictions on external financing and reduces financing cost.
In China, the government and business groups control a large number of listed
firms through the pyramidal ownership structure. However, they have different
incentives for creating this pyramid structure within their organizations (Fan et al.,
2005; Zhu, 2006). SOEs utilize the pyramidal ownership structure to mitigate
government interference on listed firms and allow them more flexibility in operation
under a free market. For NOSEs, ultimate shareholders use the pyramid to control
those firms with less cash flow and therefore create an internal capital market within
2
the pyramid (Bianco and Casavola, 1999; Attig et al., 2003; Fan et al., 2005). This
paper attempts to investigate the following issues: What role does the pyramidal
ownership structure play on a firm’s investment decision? Will more layers in the
pyramid result in more capital investments due to lower financing cost? What is the
impact of increasing layers in the pyramid on a firm’s performance?
In order to study the effect of pyramidal ownership structure on capital
investment, this paper uses two methods: the cash flow-investment sensitivity based
on Fazzari et al. (1988) and direct measurement of overinvestment based on
Richardson (2006). Our evidence indicates that as the corporate pyramid increases,
both capital investment and overinvestment decrease. That is, overinvestment is
negatively related to the pyramid measured by the layers of control chain, and the
negative relation exists for both SOEs and NSOEs. Specifically, pyramiding reduces
the likelihood of overinvestment and improves investment efficiency. We also find
that the economic influences of pyramidal ownership structure on accounting
performance of SOEs and NSOEs are not the same. For SOEs, the pyramid is
positively related to accounting performance, not only due to the decrease in
government interference but also due to the effect of restricting overinvestment. The
evidence can be attributed to the fact that overinvestment may waste resources and
reduce return, while the pyramid may mitigate overinvestment and enhance firm
performance. For NSOEs, the pyramid structure may hurt accounting performance
because the agency cost may dominate the positive effect of pyramid on
overinvestment.
3
The contributions of this paper include the following two aspects: First, we
investigate the role of internal capital market within the pyramid and whether the
pyramid can restrict the overinvestment. In addition, we study the relationship
between corporate pyramid and overinvestment of SOEs and NSOEs. Second, we
examine the influence of pyramidal ownership structure on firm performance for
SOEs and NSOEs due to their different incentives of creating corporate pyramids.
Our findings offer some evidences related to the successful reform of SOEs in China,
and the efficiency for market operation and government interference.
The rest of the paper is organized as follows: Section 2 reviews the related
literatures. Section 3 presents our hypothesis and theory. The empirical results are
reported in Section 4 through Section 5. Finally, we conclude this paper in Section 6.
2. Literature Review
Concerning the internal capital market, Lee et al. (2004) propose two competing
views: First, the value added view suggests that internal capital market can enhance
firm value. Second, the tunneling view suggests that internal capital market can hinder
capital allocation efficiency.
The value-added view claims that the headquarters of a business group can
allocate group resources most to its member firms with the most growth opportunities
through the formation of an internal capital market, in which internally generated cash
flows can be effectively pooled among member firms or different business divisions
(Lee et al., 2004). In an ideal circumstance, the internal capital market of a diversified
firm would allow it to fund profitable projects that the external capital market would
4
not be able to finance due to information asymmetry and agency costs. As a result the
segments of diversified firms are supposed to be more efficient in their financial
decisions compared with the equivalent stand-alone firms (Iwaisako and Kobayash,
2001). If firms have limited funds available for investment because external funds are
more expensive than internal funds, an efficient internal capital market allocates these
funds to maximize shareholder wealth (Shin and Stulz, 1998). The internal capital
market’s efficiency increases if it is used for “winner picking,” where capital is
relocated for financing operations that have a solid growth prospect but will be
underfunded if operated as a standalone unit (Williamson, 1975; Stein, 1997).
This tunneling view (Johnson et al., 2000) predicts that the resource transfer from
a healthy member firm to a weaker one and claims that the internal capital allocation
tends to be inefficient (Lee et al., 2004). If a firm’s management pursues its own
objectives at the expense of shareholder wealth, it might use the firm’s internal capital
market to finance the projects with negative net present values (NPV). It might
subsidize losing divisions. Moreover, the internal capital market could fail because
each division is treated as a standalone firm that relies mostly on its own cash flow to
finance its projects. Divisional managers may expend substantial resources in
rent-seeking and internal politics, thereby distort the allocation of resources and create
deadweight costs. Often, the divisional managers who benefit the most from
expending resources affect the allocation of funds of the weakest division. When
resources are allocated within a firm in such a way that the most profitable projects do
not have priority, the benefit of having an internal capital market disappears (Shin and
5
Stulz, 1998).
Given the mixed viewpoint, the issue of whether diversified firms allocate capital
efficiently or poorly remains unsolved. These two contradicting views are not only
suited for diversified firms, but also for those firms with their internal capital market
connected by pyramid structures because the pyramid structure is similar to the
internal structure in a diversified firm.
The ultimate shareholders can build their empire through a “pyramid” with
limited costs. This phenomenon is more significant in countries and regions with
weaker laws and undeveloped economies (La Porta et al., 1999; Claessens et al.,
2000). Attig et al. (2003) point out that pyramid structure is popular around the world
due to the limited responsibility and private benefit of control right. The risk of
ultimate shareholders can be diversified through the pyramid structure and the length
of pyramid layers can be changed through the pyramid structure to make high risk
projects further away from the top of the pyramid. Moreover, the ultimate shareholder
can gain the private benefit of the divergence between control rights and cash flow
rights (Attig et al., 2003). Pyramid structure is a form of organizational structure and
for the purpose of allocating the resources within the group more efficiently. The
pyramid structure enables the ultimate shareholder to grasp a larger number of assets
with limited capital. Bianco and Casavola (1999) point out that the pyramid structure
makes the obtaining external capital for the firms within the group more convenient
and the internal capital market created by the pyramid can help to launch investment
projects. Khanna and Palepu (2000) suggest that the reason for the pyramid structure
6
is that the pyramid can compensate for the market incompletion. When the market is
undeveloped, the pyramid group produces the capital market, managerial market and
intangible assets market within the pyramid. The internal capital market can lead to
convenience for the firms within the group and reduce the reliance on external
financing (Williamson, 1975; Stein, 1997). When the external financing market is
more constrained, the internal capital market is more convenient (Almeida and
Wolfenzon, 2006). Fan et al. (2005) suggest that the reason for creating a pyramidal
ownership structure in China is different for SOEs and NSOEs. For SOEs, the
pyramid is created for less government intervention, more flexible operation, and
developing the market mechanism. While for NOSEs, the pyramidal ownership
structure is to create an internal capital market.
3. Hypothesis
Figure 1 provides a theoretical framework of the relationships among pyramid,
capital investment and firm performance.
(Insert Figure 1 about here)
Concerning the influence of pyramid on capital investment, a longer pyramid
may restrict the investment of listed firms, especially the overinvestment, for both
SOEs and NOSEs.
For SOEs, a shorter pyramid means more direct intervention from the
government. The executives of listed SOEs are like to be appointed by the
government. For meeting their political performance and avoiding their future
political career in jeopardy, those government officers-executives have incentives to
7
build and expand their empires, which may lead to overinvestment. If the purpose of
creating a pyramid is to reduce the influence by the government and increase the
managerial control of the firm, the incentive for creating an empire is much lower
than that of the executive directly appointed by the government. Therefore, as the
increase of pyramid in SOEs, capital investment may be reduced and is less likely to
be overinvested.
For NOSEs, creating a pyramid aims to create an internal capital market to reduce
the reliance on external financing (Stein, 1997). The internal capital market bonding
in all related firms is convenient for resource and capital transfer among those firms to
meet the requirement of strategic development of ultimate shareholders. The decision
for NSOEs, to build an empire is not controlled or decided by the management of the
listed firms. Thus, the listed firm cannot expand its size through an internal capital
market. The ultimate shareholders may deprive resources from their listed firm
whenever they need capital. This behavior is known as “tunneling”. Thus, investments
in listed firms are restricted by the need of ultimate shareholders. Overinvestment is
easier to control through a longer pyramid.
Hypothesis 1: The Pyramid will restrict the capital investment of listed firms
and improve the efficiency of investment for both SOEs and NSOEs.
Capital investments are necessary for firms’ persistent operation and future
growth. Profitable long term investment enhances firm’s future cash flows; therefore
capital investment requirement for expansion is positively related to firm performance.
However, firm’s management, in East Asian the controlling shareholders or ultimate
8
shareholders, may pursue their own objectives at the expense of other shareholders’
wealth, and they might use the firm’s internal capital market to finance the projects
with negative net present value (NPV). It may subsidize losing divisions or engage in
overinvestment activities. When resources are allocated within a firm in such a way
that capital is expanded on those projects, overinvestment will reduce firm
performance due to resource waste and further lowering the capability to financing.
Thereby, overinvestment will lower the accounting performance.
If firms are constrained by external financing, maybe it lacks of necessary
resources to expand or financing for some profitable projects. Then it will lose its
opportunity which maybe captured by other firms. If this situation is continues, fierce
competitions in the market will bring the firm to smaller and smaller market and
profits. And if the bad cycle is repeated, then firm performance will be lower and
lower. Thus, underinvestment will also cause lower accounting performance.
In all, investment for firm expansion is necessary; however it should not be
overinvested or underinvested since each situation will both lead to a bad cycle in
current and future operations.
Hypothesis 2: Overinvestment (underinvestment) will lower the accounting
performance.
For SOEs, as the increase of a pyramid, less government direct intervention is
found in listed firms (Fan et al., 2005). The pyramid will reduce the incentive and
power of the managers of SOEs for building the empire and reduce the likelihood of
overinvestment. That is, listed firms can invest on other profitable projects to earn
higher profits and use the capital more efficiently. Therefore, more layers may reduce
9
overinvestment and enhance firm performance.
For NOSEs, as the pyramid increases, the internal capital market will be more
powerful (Fan et al., 2005; Stein, 1997); therefore capital and resource transfer will be
easier. For the need of ultimate shareholders, the investment of their listed firm may
be lower. If abandoned investment is the necessary investment for expansion and
enhancing competitiveness, which may reduce the competitiveness in the product
market, and further reduce the profitability and firm performance.
In another perspective, the incentive to build a pyramid for SOEs is to reduce
government direct interference, to let firms operation under market mechanism and
more flexible. More government interference will increase social burden to firms and
result in a negative effect on firm performance. Operation under a free market will
avoid government intervention, which will enhance the value and profitability of
firms.
For NSOEs, creating pyramid can create an internal capital market to reduce the
reliance on external financing in which the cost of capital is higher. In theory, this
structure will be beneficial for a listed firm, but actually under this structure, ultimate
shareholders can tunnel the firm in a more invisible manner, especially in the country
or region with weaker law or undeveloped economy. The influence of pyramid on
NSOEs depends on the efficiency of the internal capital market and the tunneling
effect.
For all firms, both SOEs and NSOEs, a problem related to the pyramid structure
is a multi-layer agency problem. With the increase of layers, more agency problems
10
will arise. Agent in one layer is also a principal in another lower layer, and each
principal-agency relation will increase agent costs and reduce efficiency. Meanwhile,
the multi-layer principal-agent problem will lead to a lower flow of information and
higher information cost. These will facilitate agents in each layer to purse private
benefits using their information advantages. To reduce the agency problem, the
monitoring cost will increase. Moreover, the incentive problem is more severe in a
multi-layer principal-agency relation. Firm performance tends to be worse if there is
no incentive mechanism.
Although the layer of pyramid in SOEs is usually longer, a proportion of the
agency costs, such as monitoring costs, are assumed by government. For SOEs, we
assume that the positive effect of less government intervention and social burden will
be more significant than the multi-layer principal-agency problem. That is, the
influence of the pyramid on firm performance is positive for SOEs. While for NOSEs,
we hypothesize that they have more severe agent problem and the agent cost exceeds
the benefit from the internal capital market. As a consequence, the influence of
pyramid on performance will be negative.
Hypothesis 3: For SOEs, pyramid is positively related to performance, while
for NSOEs, the relation is negative.
4. Variables and Data
4.1. Variable Definition
4.1.1. Pyramidal ownership Structure, Capital Investment and Performance
The pyramidal layer (CHAIN) is the number of layers between the listed
11
company and the ultimate shareholder as defined by Fan et al., (2005) and Zhu
(2006).
Capital investment is a proxy using the following two different measures. I1 is the
increase of long term assets standardized by the beginning assets obtained from the
balance sheet. I2 is the cash purchase in long-term equity investment, debt investment,
fixed assets, intangible assets and other assets, then minus the cash flow from selling
of fixed assets, intangible assets and other assets standardized by the beginning assets.
The above data are obtained from the cash flow statement and. balance sheet.
Performance is the accounting performance measured by the net income and
income before extraordinary items. ROA is the net income divided by the average
assets. ROE is the net income divided by the average equity. EXBIOA is the income
before extraordinary items divided by the average assets, and EBXIOE is the income
before extraordinary items divided by the average equity.
4.1.2. Other Variables
Control variables include: It-1 is the capital investment of the prior year; CF is the
beginning cash flow from operation divided by beginning assets; LEV, the debt ratio
at the beginning of year; Size is the nature log of beginning assets. Q is prior year
ending Tobin’s Q, which is calculated as the market value of assets divided by book
value; 1 Sale is the prior year’s sales revenue divided by beginning assets; Ret is the
prior year’s market return; Age is the time span from IPO year; FCF, the free cash
1 In China A-share market, before 2005 not all the stock is circulated in the market. Therefore the market value of
equity is not easy to get. Researchers often compute the market value of equity as follow, market value of equity is
the market value of circulated stock plus the book value of un-circulated stock, therefore the Tobin’s Q is
calculated as: (market value of circulated stock + book value of un-circulated stock + book value of debt)/book
value of asset.
12
flow, equals to cash flow from operation minus expected capital investment derived
from Richardson (2006) expected investment model; 2 State, dummy variable, 1 is a
SOE and 0 otherwise; V, the total voting rights of the ultimate shareholder in a listed
company; CV is the deviation of cash flow right from voting right, measured by the
cash flow right divided by voting right; Years is the yearly dummy (5 dummy
variables for six year samples); Inds is the industrial dummy (after dropping the
finance industry, there are 11 dummy variables for 12 industries).
4.2. Sample
We choose all listed firms in China’s A-share market from 2001 to 2006. 3 Then,
we exclude those firms (1) without the information of ultimate shareholders, (2) in
finance industry, (3) not listed in the previous year, (4) issues other kind of shares,
like B/H/S/ADR, 4 (5) with leverage ratio greater than 5 in the prior year. 5 Finally, we
have 6,213 firm-year observations. The samples by year are reported in Table 1.
(Insert Table 1 about here)
The information of ultimate shareholder is collected manually from the annual
report of all listed firms. Other data are obtained from the Wind database.
In order to avoid the influence of outliers, we winsorize observations of the top
2
Since we use two capital investment proxies, we calculate the FCF when running regressions for each
corresponding investment proxy.
3
Since 2001 accounting principle in China changed a lot, to be consistent in data we begin sampling in 2001. And
China adopt the new accounting principle since 2007, again to be consistent in data we end sampling in 2006.
The computation of Tobin’ Q for those companies is too much complicated, and also those firms are facing
different legal environment and investors from those just issue A shares, financing constrains are different.
Therefore, to avoid the inconsistent of the Q for those firms, we drop those samples.
4
13
and bottom 1% for capital investment, I1 and I2, and top and bottom 2% for
accounting performance, ROA, ROE, EBXIOA, and EBXIOE. 6
5. Empirical Test
5.1. Descriptive statistics
Table 2 shows the statistics for those regression variables. Average profitability
proxies of sample firms are not satisfied, since it’s only around 3%. Average capital
investment is about 6% for the measure based on balance sheet, while this rate is 11%
for the proxy based on cash disbursement. These two proxies are much different
therefore we should use two proxies to give robustness tests. Pyramid layers for
sample firms are 2.4, which mean that there is at least one firm existing between
ultimate shareholder and listed firm. For some samples ultimate shareholder control
listed firm directly, while there are 7 firms from the apex to the bottom of the
pyramid.
(Insert Table 2 about here)
5.2. Correlation analysis
Table 3 shows the correlation coefficients for pyramid, capital investment and
firm performance for SOEs and NSOEs, respectively. Accounting performance is
significantly positively related to capital investment since more investments will lead
to larger economies of scale and more productive assets. This positive relation holds
true for SOEs and NSOEs.
Capital investments are negatively related with pyramid, CHAIN, both for SOEs
6
1% or 2% does not affect our results. In the robust check, we winsorize by 1%, and the results remain the same.
14
and NSOEs, for Pearson and Spearman correlation coefficients. Hypothesis 1 is
partially supported, indicating that pyramid may constrain capital investments.
While for the relation between pyramid and accounting performance, it shows
different signs for SOEs and NSOEs. For SOEs, this relation is significantly positive,
which indicating that increasing layers in pyramid leads to better performance. While
it is negative and significant for NSOEs, increasing layers in pyramid results in lower
profitability. Hypothesis 2 is supported by the correlation analysis.
(Insert Table 3 about here)
5.3. Pyramid Structure and Capital Investment
Correlation analysis just shows the relation between pyramid and investment, but
not show whether the pyramid can restrict the overinvestment and improve
investment efficiency. Therefore, here we use two methods to investigate the
influence of pyramid on investment activities, the investment-cash flow sensitivity
and direct measurement of overinvestment (underinvestment).
5.3.1. Investment-Cash Flow Sensitivity Analysis
The first method is the investment-cash flow sensitivity used first by Fazzari et al.
(1988). Higher sensitivity will indicate an overinvestment tendency, the free cash flow
hypothesis (Jenson, 1986; Vogt, 1994). And the sensitivity of investment-cash flow
can be a proxy for the inefficient investment (Biddle and Hilary, 2006). If the pyramid
can reduce investment, especially those over investments by listed firms, then it will
lower the sensitivity of investment-cash flow. Table 4 shows the influence of pyramid
15
on investment- cash flow sensitivity using model based on Fazzari et al. (1988). The
first two columns show regression results using the capital investment proxy based on
balance sheet, the rest columns are results for capital investment proxy based on cash
flow statement.
(Insert Table 4 about here)
Corporate pyramid as measured by CHAIN is significantly negatively related to
capital investment, which means that longer pyramid restricts the capital investment
of listing firms. Our results suggest that pyramiding does affect capital investment.
The sensitivity of investment-cash flow can be a proxy of the inefficient investment
(Biddle and Hilary, 2006). If corporate pyramid may reduce investment, especially
those firms with overinvestments, then it may lower the sensitivity of investment-cash
flow. Therefore, we investigate the function of pyramid on investment based on the
sensitivity of investment-cash flow.
After controlling for other influences, the cross term of CHAIN and CF is
negatively related with capital investment for all samples, which means pyramid
layers can reduce the cash flow-investment sensitivity that is the proxy for investment
inefficiency. Pyramid can restrict the overinvestment activities for both capital
investment proxies, improving the investment efficiency proxy by the cash
flow-investment sensitivity (Biddle and Hilary, 2006).
We further divide samples into SOEs and NSOEs, finding that the restrictive role
of pyramid on capital investment is significant both for SOEs and NSOEs.
(Insert Table 5 about here)
16
The results in Table 5 indicate that as the pyramid increases, the capital
investment decreases, negatively related to pyramid, and the negative relation holds
for both SOEs and NSOEs. Hypothesis 1 is supported. Specifically, it shows that the
internal capital market can constrain the capital investment of listed firms and
improve investment efficiency.
5.3.2. Overinvestment Measurement
The
second
method
is
the
direct
measurement
of
overinvestment
(underinvestment) by Richardson (2006). Table 6 presents the results of the relation
between pyramid and overinvestment (underinvestment) which is derived from
expected investment model by Richardson (2006). 7 The first three columns show
regression results using the first proxy for investment and the last three columns are
results for the second measurement for investment. Under each investment proxy, we
regress for total samples, and SOEs and NSOEs sub-samples.
(Insert Table 6 about here)
Controlling for the free cash flow problem during investment proposed by Jenson
(1986) and the influence of ownership structure, 8 we find that as the pyramid
increases, the overinvestment declines, and negatively related to pyramid, and the
negative relation holds true for both SOEs and NSOEs. Five out of six coefficients for
CHAIN are significantly negative. In all, Hypothesis 1 is supported, the internal
7
To simplify, we directly use the residuals from Richardson’s (2006) model to proxy for overinvestment
(underinvestment).
8
We use FCF to proxy for this effect.
17
capital market will constrain the capital investment of listed firms, restricting the
overinvestment and improving investment efficiency.
5.3.3．Pyramid, Capital Investment and Performance
We then investigate the relation among pyramid, capital investment and
accounting performance. Since the internal capital market created by the pyramid
structure may limit the capital investment of listed firms, the operating performance is
affected.
In Table 7, we use I1 as the proxy of capital investment and OverI1 is the residuals
from Richardson’s (2006) model based on I1. The first four columns are the results for
ROA, and the rest four columns are for ROE which is another proxy for firm
performance 9 . For each performance proxy, we first test the influence of capital
investment, and then we examine the effect of overinvestment (underinvestment) for
hypotheses 2. In the next regression, we test the agency problem associated with
pyramid, and in the final regression we test the net effect of pyramid on accounting
performance through the restriction on overinvestment (underinvestment) and agency
problem for hypotheses 3. When testing the influence of pyramid on performance, we
first use a dummy variable to estimate the different influences of pyramid for SOEs
and NSOEs.
(Insert Table 7 about here)
For both performance proxies, Table 7 shows that accounting performance is
9
Actually, regression results for EBXIOA and EBXIOE are the same as ROA and ROE. To be concise, we just
show regressions based on ROA and ROE.
18
positively related to capital investment since more investments lead to larger
economies of scale and more productive assets, which we call “investment effect”.
However, overinvestment may input resources on projects with a negative NPV,
causing inefficiency in investment therefore leading to poor performance. While
underinvestment also hurts future growth and value-added activities, leading to lower
market share and weaker performance. This influence we call “overinvestment effect”.
Thus, using the value of the residuals from Richardson’s (2006) expected investment
model, which directly measure the scale of overinvestment or underinvestment, we
find that more overinvestment (underinvestment) will have a negative influence on
accounting performance indicating by the significantly negative coefficients for
OverI1, in 0.01 levels. Hypothesis 2 is supported. R2 for this regression is 27.94%
based on ROA and 14.68% based on ROE, and the incremental explanation for
“overinvestment” on performance is 2.98% and 2.46% respectively.
Since the internal capital market created by the pyramidal ownership structure
may limit the overinvestment of listed firms and improve the investment efficiency,
which has been proved in Table 4 and Table 6. Thus, we show that pyramid is
beneficial for firm performance since it reduces overinvestment.
In the third column, we investigate the agency costs associated with the creation
of internal capital market, which is one of the negative effects of pyramiding. We find
that for all firms, increasing layers of the pyramid results in more severe agency
problem which has a negative influence on performance. While for SOEs, longer
pyramid or control chain may alleviate the agency problem. For less government
19
intervention, the benefits related to more layers in pyramid dominate the cost, leading
to a positive effect on performance. Coefficients for CHAIN are significantly negative,
at 1% levels, and for STATECHAIN are positive. We also use Wald statistics to
measure the net influence of pyramid for SOEs, although it is not significant in the
ROA regression.
In the fourth column, we combine the benefits of pyramid and its cost due to
agency problem to measure the net effect of pyramid. We find that after controlling
for other factors, “investment effect” and “overinvestment effect” on performance, the
relations between pyramid and accounting performance remain different for SOEs and
NSOEs. The net effect for NSOEs remains negative. That is, agency costs associated
with pyramid dominate the benefits from reducing overinvestment, which is
confirmed by the significantly negative coefficients for CHAIN. While for SOEs, less
government intervention dominates agency costs. Combined with the benefits of
restricting overinvestment, increasing pyramid for SOEs is positively related to firm
performance. Significantly positive coefficients for STATECHAIN and significant
Wald statistics indicate that coefficients for (CHAIN+STATECHAIN) are
significantly positive. Thus, the hypothesis 3 is supported. Moreover, controlling for
the benefit of pyramid, which is “restricting overinvestment”, can bring additional
7.66% explanation for the regression of ROA and 5.56% for ROE.
Furthermore, we run regressions of each subsample in Table 8 for robust tests.
The accounting performance proxy in Table 8 is ROA, and we also use ROE,
20
EBXIOE, and EBXIOA as robust tests.10 Results in Table 8 are basically the same as
shown in Table 7. The effects of pyramid on accounting performance are not the
same for SOEs and NSOEs after controlling for the investment effect, the
overinvestment effect and the agency effect. For SOEs, with the increase of pyramid,
performance tends to be better. Coefficients for CHAIN are significantly positive,
which consistent with the cross-term results. For SOEs, the pyramid is beneficial for
firm performance, not only for the less government intervention, but also for the
effect on restricting overinvestment.
However for NSOEs, the pyramid negatively affects accounting performance
although it has a positive influence on overinvestment. The above positive effect is
offset by higher agency costs related to the pyramid, leading to a negative relation
between performance and pyramid. That is, pyramid has a negative effect on NSOEs,
which suggests more layers in the pyramid inducing lower performance.
In summary, the results in Table 7 and Table 8 using the first investment measure
are consistent with hypothesis 2 and hypothesis 3.
5.4. Sensitivity Analysis
Using another proxy for the measurement of capital investment, I2, which is
based on cash flow statement, we find that regression results in Table 9 are consistent
with what we find in Table 7 and Table 8. Robust tests using another capital
10
To simplify, we don’t report these regressions. However, the results are available
upon request.
21
investment measurement in Table 9 show that: the relationships between pyramid and
performance of the listing firm are not the same for SOEs and NOSEs. For SOEs, the
pyramid is positively related to performance because as the pyramid increases,
overinvestment is less likely and results in a positive effect on firm performance.
While for NOSEs, the pyramiding also reduces capital investment as the pyramid
increases. But more agency problem accompanied with lower the performance.
Since extraordinary items can be used to manipulate net income, using ROA or
ROE to measure accounting performance may be distorted when extraordinary items
are included in the new income. To use a better measure of earnings quality, we
employ income before extraordinary items to estimate ROA and ROE as defined as
EBXIOA and EBXIOE. Our regression results are basically the same as previous
findings. Specifically, increasing the layers of control chain in the pyramid, the
performance of the SOE (NSOE) improves (deteriorate). For SOEs, as the pyramid
increases, accounting performance improves because increasing the layers of control
chain in the pyramid may reduce overinvestment due to less government intervention.
For NSOEs, increasing the layers in the pyramid may reduce capital investment, but
higher agency costs seem to offset the benefit of less overinvestment and result in
weaker performance.
(Insert Table 9 about here)
6. Conclusions
A business group structured by the corporate pyramid enables the ultimate
shareholder to control its portfolio firms in the pyramid with less cash requirement.
22
Thus, the business group creates an internal capital market and manages capital and
recourses allocations within the corporate pyramid. It will help business group
alleviate restrictions of external financing and reduce financing cost. Stein (1997)
argues that using a pyramid structure to build an internal capital market may mitigate
the reliance on external financing. When the external financing market has greater
constraints, the internal capital market becomes more convenient (Almeida and
Wolfenzon, 2006). This paper attempts to answer the following questions: How does
the internal capital market influence the capital investment? What is the economic
consequence of the pyramidal ownership structure? Do corporate pyramids affect the
performance of listed firms?
We investigate the relationships among capital investment, firm performance and
pyramidal ownership structure in China. Our empirical results indicate that longer
layers in the pyramid reduce the likelihood of overinvestment. The negative relations
between pyramidal ownership structure and overinvestment exist for SOEs and
NSOEs. Since the longer layers in the pyramid may restrict overinvestment and
improve investment efficiency, result in better firm performance. Furthermore,
different incentives for creating a pyramid structure may induce different effects on
performance for SOEs and NSOEs. For SOEs, the creation of corporate pyramid
results in less government interferences and offers more freedom for firms to operate
in a free market. Therefore, pyramiding may enhance the profitability of SOEs.
While for NSOEs, although corporate pyramid may impose more restrictions on
overinvestment, higher agency costs dominate the positive effect of pyramid on
23
overinvestment and lead to lower accounting performance for NSOEs. This paper
offers evidence of the benefits of pyramiding. Our results show that pyramid can
restrict overinvestment and improve the efficiency of capital investment.
24
References
Almeida, H. and D. Wolfenzon, 2006. A theory of pyramidal ownership and family
business group, Journal of Finance 61:2637-2680.
Attig, Najah, Yoser Gadhoum and Larry Lang. 2003. Bid-ask spread, asymmetric
information and ultimate ownership, Working Paper, The Chinese University of Hong
Kong.
Bianoco, Magda and Paola Casavola, 1999. Italian corporate governance effects on
financial structure and firm performance, European Economic Review 43, 1057-1069.
Brennan, M.J. and A. Subrahmanyam, 1996. Market microstructure and asset
pricing: on the compensation for illiquidity in stock returns, Journal of Financial
Economics 41: 441-464.
Easley, D., and M. O’Hara, 2004. Information and the cost of capital, Journal of
Finance 59, 1552-1583.
Claessens, S., S. Djankov, and L. Lang, 2000. The separation of ownership and
control in East Asian Corporations, Journal of Financial Economics 58, 81-112.
Chen, Chao, and Song Zhu. The Characteristics of Ultimate Shareholder, Government
Control and the Performance of Companies in China, America Accounting
Association 2006 Conference, Washington D.C
Fan, Joseph, P.H. and T.J. Wong, 2002, Corporate ownership structure and the
informativeness of accounting earnings, Journal of Accounting and Economics 33,
401–425.
Fan, Joseph, P.H., T.J. Wong, and T. Zhang, 2005, The emergence of corporate
pyramids in China, Working Paper, The Chinese University of Hong.
KongFazzari, S. M., R.G. Hubbard, and B. C., Peterson, 1988. Financing constrains
and Corporate Investment. Brookings Papers on Economic Activity, 201-219.
Iwaisako, Tokuo and Gaku Kobayashi, 2001, Internal Capital Markets of the Japanese
Firms, Working Paper, Hitotsubashi University.
Jensen, M., 1986. Agency Cost of Free Cash Flows, Corporate Finance, and
Takeovers. American Economic Review 76, 323-329.
Johnson, Simon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer,
2000, Tunneling, American Economic Review 90, 22-27.
25
Khanna, Tarun and Tkrishna Palepu, 2000. Is group affiliation profitable in emerging
markets? An analysis of diversified Indian business groups. Journal of Finance 55,
867-893.
La Porta, R., F. Lopez-de-Silanes and A. Shleifer, 1999, Corporate ownership around
the world, Journal of Finance 54, 471-518.
Lee, Sangwoo, Kwangwoo Park, and Hyun-Han Shin, 2001.The very dark side of
internal capital markets: Evidence from diversified business groups in Korea,
Working Paper, University of Illinois.
Myers, S., and N. Majluf., 1984. Corporate financing and investment decisions when
firms have information that investors do not have. Journal of Financial Economics 13,
187-221.
Richardson, S., 2006. Over-investment of free cash flow, Review of Accounting
Studies 11, 159-189.
Stein, J.C., 1997, Internal capital markets and the competition for corporate resources,
Journal of Finance 52, 111-133.
Sudip. Datta, Ranjan D’Mello, and Mai Iskandar-Datta, 2008. Executive
compensation and internal capital market efficiency, Journal of Financial
Intermediation, forthcoming;
Shin, H., R. Stulz,, 1998. Are internal capital markets efficient? Quarterly Journal of
Economics 113, 531–552.
Vogt. S.C., 1994. The cash flow-investment relation: Evidence from U.S.
manufacturing firms, Financial Management 23, 3-20.
Williamson, Oliver E. 1975. Markets and Hierarchies: Analysis and Antitrust
Implications, New York: Free Press.
Zhu, Song, 2006. The characteristics of ultimate shareholders and informativeness of
accounting earnings, China Accounting and Finance Review 3, 1-30.
26
Figure 1
Theoretical Framework on the Relationships among Pyramid, Capital
Investment and Firm Performance
Market Operation
Capital
Investment
Internal Capital Market
Pyramid
Government interference
Underinvestment
Overinvestment
reduces return
reduces return
Performance
Agency
Problem
Graph 1 Relation
among
Pyramid, Capital Investment and Performance
27
Table 1
Number of Firm-Year Observations by SOEs and NSOEs: 2001-2006
2001
2002
2003
2004
2005
2006
Total
Total
861
935
1,010
1,073
1,164
1,170
6,213
SOEs
716
737
759
760
793
761
4,526
NSOEs
145
198
251
313
371
409
1,687
Table 2
Descriptive Statistics for Variables
ROA, the net income divided by the average assets, and EBXIOA, net income before
extraordinary items divided by the average assets; ROE, the net income divided by the average
equity, and EBXIOA, net income before extraordinary items divided by the average equity; these
four performance proxies have been winsorized for top and bottom 2%. I1 is the increase of long
term assets standardized by the beginning assets; I2 is the cash purchase in long-term equity
investment, debt investment, fixed assets, intangible assets and other assets, then minus the cash
flow from selling of fixed assets, intangible assets and other assets, standardized by the beginning
assets; these two proxies have been winsorized for top and bottom 1%. CHAIN is the number of
layers from listed company to the ultimate shareholder. CF is the beginning cash flow from
operation divided by beginning assets; LEV, the debt ratio at the beginning of year; Size is the
nature log of beginning assets. Q is Tobin’s Q, which is calculated as the market value of assets
divided by book value; Sale is the prior year’s sales revenue divided by beginning assets; Age is
the time span from IPO year.
N
Mean
Sd
Min
Median
Max
ROA
6213
0.0170
0.0640
-0.2344
0.0244
0.1284
EBXIOA
6213
0.0140
0.0585
-0.1999
0.0202
0.1236
ROE
6213
0.0321
0.1656
-0.7163
0.0568
0.3214
EBXIOE
6213
0.0261
0.1512
-0.6230
0.0475
0.3103
I1
6213
0.0609
0.1324
-0.3220
0.0366
0.5589
I2
6213
0.1125
0.1175
-0.0318
0.0774
0.5816
CHAIN
6213
2.3544
0.7551
1
2
7
CF
6213
0.0453
0.0873
-1.3058
0.0455
1.0192
Lev
6213
0.4821
0.2583
0.0081
0.4704
4.3427
Size
6213
21.0119
0.8674
17.4965
20.9520
25.7343
Q
6213
1.4632
0.5832
0.7030
1.2857
11.7343
Sale
6213
0.5749
0.4549
0
0.4529
5.1237
Age
6213
6.0900
3.1414
1
6
16
28
Table 3
Correlation Coefficients of Performance, Capital Investment and Corporate
Pyramid
ROA, the net income divided by the average assets, and EBXIOA, net income before
extraordinary items divided by the average assets; ROE, the net income divided by the average
equity, and EBXIOE, net income before extraordinary items divided by the average equity; I1 is
the increase of long term assets standardized by the beginning assets; I2 is the cash purchase in
long-term equity investment, debt investment, fixed assets, intangible assets and other assets, then
minus the cash flow from selling of fixed assets, intangible assets and other assets, standardized by
the beginning assets; CHAIN is the number of layers from listed company to the ultimate
shareholder. Above the diagonal are Spearman correlations, and under the diagonal are Pearson
correlations. ***, **, and * denote significant at the 0.01, 0.05, and 0.10 level, respectively.
Panel A-SOEs
ROA
EBXIOA
ROE
EBXIOE
I1
I2
CHAIN
ROA
0.3537***
0.4136***
0.0308**
EBXIOA
0.3733***
0.4189***
0.0262*
ROE
0.3257***
0.3505***
0.0307**
EBXIOE
0.3423***
0.3576***
0.0344**
I1
0.2988***
0.3267***
0.2475***
0.2693***
-0.0343**
I2
0.2996***
0.3138***
0.2395***
0.2566***
-0.0270*
CHAIN
0.0267*
0.0190
0.0370**
0.0407***
-0.0418***
-0.0261*
EBXIOA
ROE
EBXIOE
I1
I2
CHAIN
ROA
0.3875***
0.4307***
-0.2192***
EBXIOA
0.4204***
0.4447***
-0.2294***
ROE
0.2408***
0.2604***
-0.1521***
EBXIOE
0.2695***
0.2784***
-0.1683***
Panel B-NSOEs
ROA
I1
0.3625***
0.3962***
0.2069***
0.2348***
-0.1292***
I2
0.3115***
0.3254***
0.2015***
0.2071***
-0.1369***
CHAIN
-0.1752***
-0.1860***
-0.1343***
-0.1374***
-0.0826***
-0.1152***
29
Table 4
Pyramid, Internal Capital Market and Capital Investment
I1 is the increase of long term assets standardized by the beginning assets; I2 is the cash purchase
in long-term equity investment, debt investment, fixed assets, intangible assets and other assets,
then minus the cash flow from selling of fixed assets, intangible assets and other assets,
standardized by the beginning assets; CF is the beginning cash flow from operation divided by
beginning assets; CHAIN is the number of layers from listed company to the ultimate shareholder;
CHAINCF is the cross-term of CHAIN and CF. Control variables include: LEV, the debt ratio at
the beginning of year; Size is the nature log of beginning assets. Q is Tobin’s Q, which is
calculated as the market value of assets divided by book value; Sale is the prior year’s sales
revenue divided by beginning assets; Age is the time span from IPO year; Years, 5 dummy
variables for six year samples; Inds, after dropping the finance industry, 11 dummy variables for
12 industries. The White–adjusted t statistics considering the heteroscedasticity are in the
parentheses. ***, **, and * denote significant at the 0.01, 0.05, and 0.10 level, respectively.
I = α + β 1C F + β 2 C h a i n C F + β 3 L e v + β 4 S i z e + β 5 Q + β 6 S a l e + β 7 A g e
+ β i ∑ Years + β
j
∑
In d s + ε
I1
CF
CHAIN
I2
0.1255
0.3256
0.1664
0.3744
(4.66)***
(5.11)***
(8.09)***
(7.09)***
-0.0055
-0.0043
(-2.57)**
(-2.33)**
CHAINCF
-0.0818
-0.0851
(-3.31)***
(-4.15)***
Controls
Control
Control
Control
Control
N
6213
6213
6213
6213
R-sq
0.1086
0.1097
0.1489
0.1510
30
Table 5
Pyramid, Internal Capital Market and Capital Investment-Robust Test
I1 is the increase of long term assets standardized by the beginning assets; I2 is the cash purchase
in long-term equity investment, debt investment, fixed assets, intangible assets and other assets,
then minus the cash flow from selling of fixed assets, intangible assets and other assets,
standardized by the beginning assets; CF is the beginning cash flow from operation divided by
beginning assets; CHAIN is the number of layers from listed company to the ultimate shareholder;
CHAINCF is the cross-term of CHAIN and CF. Control variables include: LEV, the debt ratio at
the beginning of year; Size is the nature log of beginning assets. Q is Tobin’s Q, which is
calculated as the market value of assets divided by book value; Sale is the prior year’s sales
revenue divided by beginning assets; Age is the time span from IPO year; Years, 5 dummy
variables for six year samples; Inds, after dropping the finance industry, 11 dummy variables for
12 industries. The White–adjusted t statistics considering the heteroscedasticity are in the
parentheses. ***, **, and * denote significant at the 0.01, 0.05, and 0.10 level, respectively.
I = α + β 1C F + β 2 C h a i n C F + β 3 L e v + β 4 S i z e + β 5 Q + β 6 S a l e + β 7 A g e
+ β i ∑ Years + β
j
∑
In d s + ε
SOEs
NSOEs
I1
I2
I1
I2
0.2814
0.2868
0.3852
0.4684
(3.71)***
(4.59)***
(3.27)***
(4.71)***
-0.069
-0.0546
-0.0956
-0.1136
(-2.24)**
(-2.14)**
(-2.29)**
(-3.24)***
Controls
Control
Control
Control
Control
N
4526
4526
1687
1687
R-sq
0.0854
0.1438
0.1679
0.2101
CF
CHAINCF
31
Table 6
Overinvestment and Pyramidal Ownership Structure
OverI1 and OverI2 are residuals from Richardson (2006) expected investment model, the
overinvestment (underinvestment) proxy for I1 and I2; I1 is the increase of long term assets
standardized by the beginning assets; I2 is the cash purchase in long-term equity investment, debt
investment, fixed assets, intangible assets and other assets, then minus the cash flow from selling
of fixed assets, intangible assets and other assets, standardized by the beginning assets; It-1 is the
capital investment of the prior year; CF is the beginning cash flow from operation divided by
beginning assets; LEV, the debt ratio at the beginning of year; Size is the nature log of beginning
assets. Q is Tobin’s Q, which is calculated as the market value of assets divided by book value;
Ret is the prior year’s market return; Age is the time span from IPO year; CHAIN is the number
of layers from listed company to the ultimate shareholder. FCF, the free cash flow, equals to cash
flow from operation this year minus expected capital investment derived from Richardson (2006)
expected investment model; State, dummy variable, 1 indicate SOEs and 0 otherwise; V, the total
voting right of ultimate shareholder in the listed companies; CV is the deviation of cash flow right
from voting right, equals to cash flow right divided by voting right; Years, 5 dummy variables for
six year samples; Inds, after dropping the finance industry, 11 dummy variables for 12 industries.
White–adjusted t statistics considering the heteroscedasticity are in the parentheses. ***, **, and *
denote significant at the 0.01, 0.05, and 0.10 level, respectively.
I t = α + β 1 I t −1 + β 2 C F + β 3 L ev + β 4 Size + β 5 Q + β 6 R e t +
β 7 A ge + β i ∑ Years + β j ∑ Inds + ε
OverI (ε ) = α + β1Chain + β2 FCF + β3V + β4CV + βi ∑Years + β j ∑ Inds + ε
OverI1
OverI2
All
SOE
NSOE
All
SOE
NSOE
-0.0068
-0.0049
-0.0094
-0.0058
-0.0015
-0.0086
(-3.00)***
(-1.82)*
(-2.09)**
(-3.08)***
(-0.68)
(-2.48)**
0.1533
0.1415
0.1795
0.1271
0.1233
0.1486
(7.41)***
(5.66)***
(4.64)***
(7.49)***
(6.23)***
(4.44)***
0.0153
0.0068
0.0514
0.0043
-0.0023
0.0652
(1.49)
(0.59)
(2.15)**
(0.51)
(-0.23)
(3.37)***
-0.0128
0.0014
-0.0303
-0.0128
0.0116
-0.0247
(-1.75)*
(0.14)
(-2.28)**
(-2.20)**
(1.49)
(-2.46)**
Years
Control
Control
Control
Control
Control
Control
Inds
Control
Control
Control
Control
Control
Control
N
6213
4526
1687
6213
4526
1687
R-sq
0.0190
0.0167
0.0381
0.0210
0.0197
0.0548
CHAIN
FCF
V
CV
32
Table 7
Pyramid, Capital Investment and Performance
ROA, the net income divided by the average assets; ROE, the net income divided by the average
equity; I1 is the increase of long term fixed assets standardized by the beginning assets; OverI1 is
the overinvestment (underinvestment) derived from Richardson (2006) expected investment model,
proxy for the scale of overinvestment (underinvestment); CHAIN is the number of layers from
listed company to the ultimate shareholder; StateChain is the product of State and Chain, and
State is a dummy variable, 1 indicate SOEs and 0 otherwise. Control variables include: V, the
total voting right of ultimate shareholder in the listed companies; CV is the deviation of cash flow
right from voting right, equals to cash flow right divided by voting right; LEV, the debt ratio at
the beginning of year; Size is the nature log of beginning assets. Q is Tobin’s Q, which is
calculated as the market value of assets divided by book value; Sale is the prior year’s sales
revenue divided by beginning assets; ROA is the accounting performance for the prior year; Age
is the time span from IPO year; Years, 5 dummy variables for six year; Inds, after dropping the
finance and real estate industry, 10 dummy variables for 11 industries. White–adjusted t statistics
considering the heteroscedasticity are in the parentheses. Wald-F is the Wald test for coefficients
of (CHIAN+STATECHAIN) =0; ***, **, and * denote significant at the 0.01, 0.05, and 0.10 level,
respectively.
R O A = α + β 1 I 1 + β 2 O v e r I 1 + β 3 C h a i n + β 4 S t a t e C h a i n + β 5 S t a t e + β 6V + β 7 C V + β 8 L e v + β 9 S i z e +
β 10Q + β 11S a le + β 12 R O A + β 13 A g e + β
i
∑
Years + β
j
∑
In d s + ε
ROA
I1
Investment
Effect
Overinvest
Effect
0.1120
(15.60)***
OverI1
ROE
Net
Effect
Investment
Effect
Overinvest
Effect
0.5114
0.5082
0.2356
1.1743
1.1668
(17.36)***
(17.36)***
(12.70)***
(15.88)***
(15.86)***
-0.4178
-0.4147
-0.9818
-0.9744
(-14.15)***
(-14.12)***
(-13.02)***
(-12.97)***
CHAIN
STATECHAIN
Agency
Effect
Agency
Effect
Net
Effect
-0.0099
-0.0083
-0.0242
-0.0207
(-4.61)***
(-4.12)***
(-3.91)***
(-3.44)***
0.0116
0.0111
0.0307
0.0295
(4.91)***
(4.94)***
(4.62)***
(4.54)***
Controls
Control
Control
Control
Control
Control
Control
Control
Control
N
6213
6213
6213
6213
6213
6213
6213
6213
0.2496
0.2794
0.2066
0.2832
0.1222
0.1468
0.0952
0.1508
R
2
2
Add-R
Wald-F
0.0298
0.0766
1.77
5.22**
0.0246
0.0556
3.22*
33
6.37**
Table 8
Pyramid, Capital Investment and Performance by State or Non-State Ultimate
Shareholders
ROA, the net income divided by the average assets; I1 is the increase of long term fixed assets
standardized by the beginning assets; OverI1 is the overinvestment (underinvestment) derived
from Richardson (2006) expected investment model, proxy for the scale of overinvestment
(underinvestment); CHAIN is the number of layers from listed company to the ultimate
shareholder; StateChain is the product of State and Chain, and State is a dummy variable, 1
indicate SOEs and 0 otherwise. Control variables include: V, the total voting right of ultimate
shareholder in the listed companies; CV is the deviation of cash flow right from voting right,
equals to cash flow right divided by voting right; LEV, the debt ratio at the beginning of year;
Size is the nature log of beginning assets. Q is Tobin’s Q, which is calculated as the market value
of assets divided by book value; Sale is the prior year’s sales revenue divided by beginning assets;
ROAt-1 is the accounting performance for the prior year; Age is the time span from IPO year;
Years, 5 dummy variables for six year; Inds, after dropping the finance and real estate industry, 10
dummy variables for 11 industries. White–adjusted t statistics considering the heteroscedasticity
are in the parentheses. ***, **, and * denote significant at the 0.01, 0.05, and 0.10 level,
respectively.
R O A = α + β 1 I 1 + β 2 O v e r I 1 + β 3 C h a i n + β 4V + β 5 C V + β 6 L e v + β 7 S i z e + β 8 Q + β 9 S a l e
+ β 10 R O A t −1 + β 11 A g e + β
i
∑
Years + β
j
∑
In d s + ε
State
I1
Investment
Effect
Overinvest
Effect
0.0987
(11.37)***
OverI1
Non-State
Net
Effect
Investment
Effect
Overinvest
Effect
0.4580
0.4590
0.1358
0.6030
0.5899
(11.90)***
(11.93)***
(9.47)***
(10.12)***
(9.92)***
-0.3765
-0.3771
-0.4844
-0.4735
(-9.81)***
(-9.83)***
(-8.28)***
(-8.10)***
CHAIN
Agency
Effect
Agency
Effect
Net
Effect
0.0022
0.0029
-0.0113
-0.0092
(1.67)*
(2.32)**
(-4.58)***
(-4.03)***
Controls
Control
Control
Control
Control
Control
Control
Control
Control
N
4526
4526
4526
4526
1687
1687
1687
1687
0.2593
0.2864
0.2190
0.2873
0.2460
0.2772
0.1995
0.2851
R
2
2
Add-R
0.0271
0.0683
0.0312
0.0856
34
Table 9
Robust Test-Cash Capital Investment and Other Performance Proxy
ROA, the net income divided by the average assets, and EBXIOA, net income before extraordinary items divided by the average assets; ROE, the net income divided by
the average equity, and EBXIOE, net income before extraordinary items divided by the average equity; I2 is the cash purchase in long-term equity investment, debt
investment, fixed assets, intangible assets and other assets, then minus the cash flow from selling of fixed assets, intangible assets and other assets, standardized by the
beginning assets. OverI2 is the overinvestment (underinvestment) derived from Richardson (2006) expected investment model, proxy for the scale of overinvestment
(underinvestment); CHAIN is the number of layers from listed company to the ultimate shareholder; StateChain is the product of State and Chain, and State is a dummy
variable, 1 indicate SOEs and 0 otherwise. Control variables include: V, the total voting right of ultimate shareholder in the listed companies; CV is the deviation of cash
flow right from voting right, equals to cash flow right divided by voting right; LEV, the debt ratio at the beginning of year; Size is the nature log of beginning assets. Q is
Tobin’s Q, which is calculated as the market value of assets divided by book value; Sale is the prior year’s sales revenue divided by beginning assets; ROAt-1 is the
accounting performance for the prior year; Age is the time span from IPO year; Years, 5 dummy variables for six year; Inds, after dropping the finance and real estate
industry, 10 dummy variables for 11 industries. White–adjusted t statistics considering the heteroscedasticity are in the parentheses. Wald-F is the Wald test for coefficients
of (CHIAN+STATECHAIN) =0; ***, **, and * denote significant at the 0.01, 0.05, and 0.10 level, respectively.
R O A = α + β 1 I 2 + β 2 O v e r I 2 + β 3 C h a i n + β 4V + β 5 C V + β 6 L e v + β 7 S i z e + β 8 Q + β 9 S a l e + β 1 0 R O A t − 1 + β 1 1 A g e + β
ROA
All
I2
OverI2
CHAIN
STATECHAIN
State
EBXIOA
Non-State
All
State
i
∑
Years + β
ROE
Non-State
All
State
j
∑
In d s + ε
EBXIOE
Non-State
All
State
Non-State
0.2654
0.2407
0.3034
0.2417
0.2152
0.2744
0.6067
0.5850
0.5735
0.5855
0.5567
0.5533
(14.94)***
(11.39)***
(7.48)***
(14.34)***
(10.25)***
(7.78)***
(14.12)***
(12.24)***
(5.42)***
(14.92)***
(10.82)***
(6.16)***
-0.1862
-0.1691
-0.2100
-0.1757
-0.1538
-0.2045
-0.4115
-0.4062
-0.3565
-0.4107
-0.3904
-0.3812
(-10.13)***
(-8.30)***
(-4.84)***
(-10.33)***
(-7.84)***
(-5.46)***
(-9.07)***
(-8.69)***
(-3.09)***
(-10.07)***
(-8.11)***
(-3.88)***
-0.0085
0.0020
-0.0094
-0.0071
0.0013
-0.0081
-0.02100
0.0070
-0.0228
-0.0181
0.0084
-0.0205
(-4.13)***
(1.65)*
(-3.97)***
(-4.15)***
(1.15)
(-4.19)***
(-3.47)***
(1.94)*
(-3.24)***
(-3.45)***
(2.51)**
(-3.35)***
0.0105
0.0085
(4.58)***
0.0281
(4.37)***
0.0259
(4.32)***
(4.51)***
Controls
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
N
6213
4526
1687
6213
4526
1687
6213
4526
1687
6213
4526
1687
R-sq
0.2531
0.2625
0.2443
0.3163
0.3332
0.3081
0.1335
0.1665
0.1085
0.1622
0.2024
0.1283
Wald-F
2.64*
1.47
4.07**
5.87**
35