Panel Threshold Effect Analysis between Capital Structure

and Operating Efficiency of Chinese Listed Companies

Chien-Chung Neih, Ph.D.*
Tamkang University
Lihong Lu
Renmin University of China

Abstract
This paper, aiming to test whether there exists an optimal debt ratio, employs panel
threshold regression model to examine threshold effects of financial leverage on
operating efficiency (and firm value) of Chinese listed firms. The findings are that
there exists triple threshold effect and the coefficients are all positive and significant
for four regimes. These results identify that the operating efficiency is positively
related to debt level. However, the positive effect decreases when debt ratio increases
since the results show that the lower the debt to asset ratio, the higher the impact
coefficient. Our result is consistent with Trade-off Theory of Capital Structure and
Agency Theory and provides an implication that financial managers should use
financial leverage wisely to improve the operating efficiency in order to maximize the
firm value and investors should refer to the debt ratio to make investment decisions.
As corporate bond market hasnt been well developed in China and firms debt
financing mainly depends on commercial banks, our result is helpful for loan
managers in banks.
Keywords: Capital structure, Financial leverage, Debt financing, Operating efficiency,
Panel threshold effect

Corresponding author: Department and Graduate Institute of Banking and Finance, Tamkang
University, Taipei, Taiwan. TEL: 886 - 2-26215656 ext. 2591. FAX: 886-2-26214755. E-mail:
niehcc@mail.tku.edu.tw.

1. Introduction
1.1 Capital Structure Theories
How can we get stable financing sources and minimize the capital cost? How can we
run our business everlastingly and maximize the shareholders wealth? Financing
strategy is always the most important decision for financial managers. Generally
speaking, sources of corporate financing mainly come from equity and debt securities.
The firms mix of different securities is know as capital structure. How to find the best
combination that maximizes the firm value is the objective of capital structure
theories.
Modigliani and Miller (1958)s capital structure irrelevance theory states that the
firms overall market value and the WACC is independent of capital structure in a
perfect market without taxation. However, the free-tax perfect market doesnt hold in
the real world. Modigliani and Miller(1963) proposed the modified Capital structure
relevance theory, which analyzed the present value of interest tax shields at the
corporate level and found that the higher the debt ratio, the higher the firm value.
Miller (1977) extends the MM model to personal as well as corporate taxes, and
introduced the Miller theory, which considered the relative tax advantage of debt
over equity.
Nevertheless, over borrowing will lead to financial distress and even bankruptcy. The
trade-off theory balances the tax advantage of borrowing against the costs of
financial distress and states there exists the optimal capital structure. The value of the
firm with borrowing equals the value if all-equity financed plus the present value of
tax shield and minus the present value of costs of financial distress. Costs of financial
distress can be broken down to bankruptcy costs and costs of financial distress short
of bankruptcy. Bankruptcy costs include direct costs such as court fees and indirect
costs reflecting the difficulty of managing a company undergoing the liquidation or
reorganization. In financial stress short of bankruptcy, conflicts of interest between
bondholders and stockholders may lead to poor operating and investment decisions.
Jensen and Meckling(1976) analyze the agency costs led by the conflicts of interest
among managers, shareholders and creditors during a financial stress and their
implication for financing policy. Jensen (1986) finds that managers of firms with
unused borrowing power and large free cash flows are more likely to undertake
low-benefit or even value-destroying mergers so higher debt ratio can reduce agency
cost and increase firm value.
Ross(1977) proposed the signaling theory, which states that Information
1

Asymmetry exists between managers and investors so the increase of debt ratio will
be taken as a positive signal to the investors and hence increase the firm value. Myers
and Majluf(1984) proposed the pecking-order theory, which states that firms use
internal financing when available and choose debt over equity when external
financing is required. After that, Lakshimi and Myers(1999) and Chirinko and
Anuja(2000) tested static tradeoff against pecking order theory.
Scholars have done a lot of research about the determinants of capital structure choice.
Kim and Sorensen (1986) find that firms with higher insider ownership have greater
debt ratios than firms with lower insider ownership. They also find that high-growth
firms use less debt rather than more debt, high-operating firms use more debt rather
than less debt, and firm size appears to be uncorrelated to the level of debt. Titman
and Wessels(1988) state that the short-term debt ratio has a negative relationship with
the firm size. Bergman and Callen (1991) states there exists negative relationship
between the ratio of intangible assets to total assets and the debt ratio. Burgman(1996)
states that the multinational corporate capital structure was influenced by political risk,
foreign exchange rate risk and high agency cost. Erwan (2001) studies the relationship
between asset liquidity and capital structure.
The empirical research on the optimal capital structure includes: Castanias (1983)
finds that ex ante default costs are large enough to induce the typical firm to hold an
optimum mix of debt and equity. Brandley, et al. (1984) developed a theoretical model
which highlights three firm-specific factors that influence the firms optimal capital
structure: the variability of firm value, the level of non-debt tax shields, and the
magnitude of the costs of financial distress. In the paper, Brandley, et al examines
cross-sectional data and finds that firm leverage ratios are related inversely to
earnings volatility. Altman(1984) compares the present value of expected bankruptcy
costs with the present value of expected tax benefits from interest payments on
leverage and concludes that the potential impact of bankruptcy costs on firm valuation
and capital structure issues is very important. Leland and Toft(1996) examine the
optimal capital structure of a firm that can choose both the amount and maturity of its
debt. They find that short term debt does not exploit complete tax benefits but reduces
agency cost compared to long term debt so the tax advantage of debt must be
balanced against bankruptcy and agency costs in determining the optimal maturity of
the capital structure. Philosophov (1999) develops a probabilistic Bayesian approach
to assess the optimal debt to equity ratio, which affects both corporate returns and
probabilities of bankruptcy to maximize the corporate value. Erik(2002) examines the
debt ratios around the tender offers, and finds that the debt ratio may have to be
2

higher than optimal as predicted by the static trade-off model, in which tax benefits
are traded off against financial distress costs in order to effectively deter takeovers.
However, there are still empirical evidences to support Capital Structure Irrelevance
Theory. Flath and Knoeber (1980) finds that cross-sectional variation in capital
structure was best explained by differences in operating risk, including that related to
the regulatory process, and not by inter-industry differences in the tax advantage to
interest, which were quite small. The general increase in interest as a percentage of
EBIT that occurred between the earlier and later periods, was strongly related to the
very great temporal increase in the tax advantage to interest that occurred between the
two periods.
1.2 Chinese Stock Market and Financing Behavior of Listed Firms
There are two stock markets in China. One is Shanghai Stock Market; the other is
Shenzhen Stock Market. They were established in 1990 and 1991 respectively. The
establishment of stock markets was the demand of the economy reform. Open market
brought severe competition and SOEs1 began to lose money. State owned banks
encountered non-performing loans problems, operational difficulties and high risks,
which cannot meet SOEs financing needs. One purpose of the establishment of stock
markets is to stimulate the reform of SOEs. Therefore, most listed firms are SOEs,
which have great influences on local economy. Having been used to get supporting
from the government, managers of SOEs have no idea about the cost of capital. This
free cash mentality is also related to the high uncertainty of transitional economy. In a
trail-and-error process of economic reform, people cannot predict future according to
their experiences. Having no clear expectations about the future, they tend to take into
consideration only current cost and benefit in decision-making.
The development of the stock market is under strict regulation of the government.
These regulations include quota restrictions and authorization system. Authorization
system means that firms can go public only with the approval from the government.
The quota system restricts the shares can be issued by a listed firm. Furthermore,
government also restricts the issuing price.2 As a result, going public is peculiar right

SOEs: State Owned Enterprises

Prior to Dec. 1996, quota system restricts the amount of shares can be issued in each province. The
local government of province chooses the SOEs to go public, and report to CSRC for approval. In order
to increase the number of firms to be listed, the local government always limits the size of the assets
going public. In Dec. 1996, the quota system changed to restrict the number of firms to be listed.
Restrictions on issuing price is based on a predetermined range of P/E, which is 15 to 20. The
authorization system ceased to be effective in 2000. Restrictions on issuing price were also canceled in
2002.
2

to lucky firms. On the other side, corporate bond market has not been well developed
which makes issuing corporate bonds very difficult for firms. All in all, Chinese firms
have limited financing sources and access to stock market is a scarce resource.
Free cash mentality, limited financing resources induce the epidemic of the thirst of
capital. We can find many symptoms of the epidemic, such as, many firms provide
rights offering shortly after IPO, a lot of the money raised in listed firms is left unused,
firms manipulate earnings in every way to meet the qualification of rights offering
and etc. Referring to the capital structure, the debt ratio of Chinese listed firms is
relatively low.
Though the long term ignores of capital cost and debt financing, Chinese listed firms
should pay more attention to their capital structure and financing behavior. With the
development of capital market, with the mature of investors, with the reform of
Chinese government, if listed firms still pay no attention the capital cost and capital
structure, they will be beaten in the open and severe competition. Chinese listed firms
should find their optimal capital structure to reduce cost of capital and maximize firm
value. Compared to equity financing, debt financing can bring tax shield to reduce the
cost of capital. Whats more, increasing of debt financing can reduce agency cost and
improve corporate governance. For Chinese investors, debt financing is a positive
signal. Thus I predict debt financing can improve firms operating efficiency.
However, on the other side, over borrowing can lead to financial distress and even
bankruptcy. Hence, firms should carefully trade off costs and benefits of debt
financing and find optimal debt ratio. This paper intends to examine the relationship
between debt financing and operating efficiency.
Chinese scholars have done some empirical research about the relationship between
the capital structure and firm value. Wang (2003) applied multiple regression analysis
and concluded that debt financing has positive impact on firm value. Nevertheless,
when I used the same method but different sample, completely different result is
found. Accordingly, I thought it is better for us to use non-linear method than linear
regression to study the relationship between capital structure and operating efficiency.
This paper applies panel threshold regression model developed by Hansen (1996) to
test whether there exists an optimal debt ratio. When debt ratio is below the threshold
value, it has a positive relationship with firms operating efficiency, which means, at
this time, managers can increase firms operating efficiency and firm value by
increasing debt ratio. Contrarily, when debt ratio is above the threshold value, a
negative relationship exists. If managers increase the debt ratio, operating efficiency
4

and firm value will decrease.

The rest of this paper is organized as follows. Section 2 describes the data resource
and variables selection. Section 3 introduces panel threshold regression model,
including empirical model, estimation method, testing process, asymptotic
distribution of threshold estimate and etc. Section 4 presents empirical results and
analysis. Section 5 concludes.

2. Data Resource and Variables Selection

This study analyzes a sample consisting of 821 A-shares listed firms traded on the
SHSE and SZSE over December 1998-December 2002 interval. It is panel data. The
total number of observations is 4105. Firms which were listed after 1998 are not
included in the sample. All financial data we uses in this paper comes from CSMAR.
We choose ROA (Return on asset) as dependent variable. As we know, Tobin Q is a
good indicator of firm value. However, Tobin q cannot well assess Chinese listed firm
value because of the peculiar features of Chinese stock market. With the exception of
a few listed firm that do not involve state ownership, all Chinese listed firms are
restructured SOEs. For the government to have an effective control over state assets
and SOEs, shares of a firm are typically split into state shares, legal-entity shares, and
tradable shares, with the restriction that state and legal-entity shares cannot be traded
publicly. State shares are those owned by the central or local government. Legal-entity
shares are those held by domestic legal entities, such as listed companies, SOEs,
banks, etc. So it is difficult for us to evaluate the market value of non-tradable shares.
Whats more, the market value of debt and replacement cost is also difficult for
valuation. Therefore, we choose the accounting indicator ROA to evaluate the firms
operating efficiency and firm value. To fix notation, let v be the ratio of return on
asset.
The independent variable is debt ratio, which is computed by book value of total
liabilities divided by book value of total assets. In this paper, debt ratio is also the
threshold variable. Let d be debt ratio.
Table 1. Descriptive summary of ROA and Debt Ratio.
25%
Minimum

percentile

75%
Medium

percentile

Maximum

Average

Debt ratio

0.005535 0.306142638 0.42795837 0.5547367

13.58267

0.453161

ROA

-13.0837 0.014648006 0.04189124 0.0674998

0.812518

0.023634

3. Model
This paper applies Panel Threshold Regression Analysis developed by Hansen (1999)
to test the threshold effect for panel data.
3.1 Threshold Autoregressive Model
Tong (1978) proposed Threshold Autoregressive model, thereafter, this non-linear
time series model becomes very popular for economic and financial research.
When we estimate Threshold Autoregressive Model, first we should test whether there
exists threshold effect. If we accept the null hypothesis, the threshold effect doesnt
exist. Again, the existence of nuisance will make the testing statistic follow
non-standard distribution, which was called Davies Problem 3 . Hansen (1999)
suggests that a bootstrap method to compute the asymptotic distribution of testing
statistics in order to test the significance of threshold effect. Furthermore, when the
null hypothesis doesnt hold, which means, the threshold effect does exist, Chan
(1993) proves that OLS estimation of threshold is super consistent, and then derive
the asymptotic distribution. However, nuisance influences this distribution and make
it non-standard. Hansen (1999) uses simulation likelihood ratio test to derive the
asymptotic distribution of testing statistic for a threshold.
Hansen (1999) proposes to use two-stage OLS method to estimate the panel threshold
model. On the first stage, for any given threshold ( ) , we compute the sum of square
errors (SSR) separately. On the second stage, we find the estimation of by

minimization of the sum of squares. At last, we can use the estimation of threshold to
estimate the coefficient for every regime and do analysis.
3.2 Threshold Model Construction
In this paper, we want to examine whether threshold effect exists between the
financial leverage and firms operating efficiency. According to Trade-off Theory of
Capital Structure, when debt ratio increases, the interest tax shield increases;
however, on the other side, costs of financial distress increase to counteract the
positive effect of debt ratio to firm value. Thus I hypothesize that there exists an
optimal debt ratio, and try to use threshold model to estimate this ratio which can help
us understand the relationship between financial leverage and operating efficiency as

Davies Problem is one that testing statistics follow non-standard distribution because of the
existence of the nuisance (Davies, 1997,1998). Afterwards, Andrews and Ploberger(1994) and
Hansen(1996) tested again to solve the problem.
6

well as help financial managers, loan managers and investors make decisions.
So we set up single threshold model as follows.
v it = i + 1 d it I ( d it ) + 2 d it I ( d it > ) + it

(1)

Where vit represents the ratio of return on assets; d it , the threshold variable,
represents the ratio of debts to assets; i , the fixed effect, represents the
heterogeneity of companies under different operating conditions; I () is the indicator
function. The errors it is assumed to be independent and identically distributed (iid)
with mean zero and finite variance 2 ; i represents different companies; t represents
different periods.
Another compact representation of (1) is to set
= (1 , 2 )
d it I (d it )
and d it =
so that (1) equals
d it I (d it > )

vit = i + d it ( ) + it

(2)

The observations are divided into two regimes depending on whether the threshold
variable d it is smaller or larger than the threshold . The regimes are distinguished
by differing regression slopes, 1 and 2 . We will use known vit and d it to
estimate the parameters , and 2 .
3.3 Estimation
Note that taking averages of (2) over the time index t produces

vit = it + d it ( ) + it
where vi =

(3)

1 T
1 T
v

=
it , and
it i T
T t =1
t =1

1 T
(
)
d
I
d

it
it

T t =1
1 T

d i ( ) = d it ( ) =
T t =1

1 T
T d it I (d it > )

t =1
Taking the difference between (2) and (3) yields
vit* = d it* ( ) + it*

(4)

where vit* = vit vi d it* ( ) = d it ( ) d i ( ) it* = it i

Let

vi*2
d i*2 ( )
i*2

*
vi* = M d i* ( ) = M
i = M
v *
d * ( )
*

iT
iT
iT
denote the stacked data and errors for an individual ,with one time period deleted.
Then let V , D ( ) and e denote the data stacked over all individuals.
v1*
d1* ( )
1*

M
M
M

*
*
*
*
*
V = vi D ( ) = d i ( ) , e = i*

M
M
M

*
*
*
n
vn
d n ( )

Use this notation, (4) is equivalent to

Vit* = Dit* ( ) + eit*

(5)

For any given , the slope coefficient can be estimated by ordinary least squares
(OLS). That is,
( ) = ( D * ( ) D * ( )) 1 D * ( )V *

(6)

The vector of regression residuals is

e* ( ) = V * D * ( ) ( )

(7)

and the sum of squared errors is

SSE1 ( ) = e* ( )e* ( ) = V * ( I D * ( )( D * ( ) D * ( )) 1 D * ( ))V *
(8)
Chan(1993) and Hansen (1999) recommend estimation of by lease squares. This
is easier to achieve by minimization of the concentrated sum of squared errors (7).
Hence the least squares estimators of is
= arg min SSE1 ( )
(9)
Once is obtained, the slope coefficient estimate is = ( ) . The residual vector is
e * = e * ( ) and residual variance

2 = 2 ( ) =

1
1
e * ( )e * ( ) =
SSE1 ( )
n(T 1)
n(T 1)

(10)

where n indexes the number of sample, T indexed the periods of sample.

3.4 Testing for a threshold
This paper hypothesizes that there exists threshold effect between the debt ratio and
operating efficiency. It is important to determine whether the threshold effect is
statistically significant. The null hypothesis and alternative hypothesis can be
8

represented as follows
H 0 : 1 = 2

H 1 : 1 2

When the null hypothesis holds, the coefficient 1 = 2 , the threshold effect doesnt
exist. When the alternative hypothesis holds, the coefficient 1 2 , the threshold
effect exists between the debt ratio and operating efficiency.
Under the null hypothesis of no threshold, the model is
vit = i + 1d it ( ) + it

(11)

After the fixed-effect transformation is made, we have

Vit* = 1d it* ( ) + eit*

(12)

~
The regression parameter is estimated by OLS, yielding estimate 1 , residuals
e~ *

and sum of squared errors

SSE 0 = ~
e * e~ *

Hansen (1999) suggests that we use the F Test Approach to test the existence of
threshold effect, and use the sup-Wald statistic to test the null hypothesis.
F = sup F ( )
(13)
( SSE0 SSE1 ( )) / 1 SSE0 SSE1 ( )
(14)
=
F ( ) =
SSE1 ( ) / n(T 1)
2
Under the null hypothesis, some coefficients (e.g. the pre-specified threshold ) do
not exist, therefore, the unisance exists. According to Davies problem (1977,1987),
the F statistic becomes non-standard distribution. Hansen (1996) shows that a
bootstrap procedure attains the first-order asymptotic distribution, so p-values
constructed from the bootstrap are asymptotically valid. Treat the regressors and
threshold variable d it as given, holding their values fixed in repeated bootstrap
samples. Take the regression residuals eit* , and group them by individual:

ei* = (ei1 , ei2 ,L, eiT ) . Treat the sample e1 , e2 ,Len

as the empirical distribution to

be used for bootstrapping. Draw a sample size of n from the empirical distribution and
use these errors to create a bootstrap sample under H 0 . Using the bootstrap sample,
estimate the model under the null (12) and alternative (4) and calculate the bootstrap
value of the likelihood ratio statistic F ( ) (14). Repeat this procedure a large
number of times and calculate the percentage of draws for which the simulated
9

statistic exceeds the actual. This is the bootstrap estimate of the asymptotic p-value
for F ( ) under H 0 . The null of no threshold effect is rejected if the p-value is
smaller than the desired critical value.
~
P = P( F ( ) > F ( ) )

(15)
~

where is the conditional mean of F ( ) > F ( ) .

3.5 Asymptotic distribution of threshold estimate

Chan (1993) and Hansen (1999) have shown that when there is a threshold effect

1 2 , is consistent for 0 , and that the asymptotic distribution is highly

non-standard. Hansen (1999) argues that the best way to form confidence intervals for
is to form the no-rejection region using the likelihood ratio statistic for tests on
. To test the hypothesis
H 0 : = 0

H 1 : 0

We construct the testing model

SSE1 ( ) SSE1 ( )
(16)
2
Hansen (1999) points out that when LR1 ( 0 ) is too large and the p-value exceeds the
LR1 ( ) =

confidence interval, the null hypothesis is rejected4. Besides, Hansen (1999) indicates
that under some specific assumptions5 and H 0 : = 0 ,
LR1 ( ) = d
as n , where is a random variable with distribution function

P( x) = (1 exp( x )) 2
2

(17)

(18)

The asymptotic p-value can be estimated under the likelihood ratio. According to the
proof of Hansen(1999), the distribution function (18) has the inverse
c( ) = 2 log(1 1 )

(19)

from which it is easy to calculate critical values. For a given asymptotic level , the
null hypothesis = 0 rejects if LR1 ( ) exceeds c( ) .

Note that the statistic (16) is testing a different hypothesis from the statistic (14) introduced in the
previous section. LR1 ( 0 ) is testing H 0 : = 0 while F ( ) is testing H 0 : 1 = 2 .
5

Refer to Hansen (1999) Appendix: Assumptions 1-8.

10

3.6 Multiple thresholds Model

If there exist triple thresholds, the model is
i

i
vit =
i
i

+ 1 d it + it L ifd it 1

+ 2 d it + it L if 1 < d it 2

(20)

+ 3 d it + it L if 2< d it 3
+ 4 d it + it L ifd it > 3

where threshold value 1 < 2 < 3 . This can be extended to multiple thresholds
model 1 , 2 , 3 , L n .

4. Empirical Results
This paper applies the threshold theory proposed by Hansen (1999) and hypothesize
that debt ration and operating efficiency have asymmetric nonlinear relationship. First
let us test whether there exist the triple thresholds effect. We test single threshold,
double threshold and triple threshold effect respectively. The models are as follows.
i + 1 d it + it L ifd it
vit =
i + 2 d it + it L ifd it >
i + 1 d it + it L ifd it 1

vit = i + 2 d it + it L if 1 < d it 2
+ d + L ifd >
3 it
it
it
2
i
i

i
vit =
i
i

+ 1 d it + it L ifd it 1

+ 2 d it + it L if 1 < d it 2
+ 3 d it + it L if 2< d it 3

, Where 1 < 2 < 3

+ 4 d it + it L ifd it > 3

The dependent variable vit represents ROA, the indicator of operating efficiency.
The independent d it represents debt ratio, in this paper d it is also threshold
variable.
This paper follows the bootstrap method to get the approximation of F statistic and
then calculate the p-value. Table 2 presents the empirical results of test for single
threshold, double threshold, and triple threshold effects. Through 200 bootstrap
replications for each of the three bootstrap tests, we find that the test for a single
threshold is highly significant with a bootstrap p-value of 0.03, the test for a double
threshold is also significant with a p-value of 0.06, and the test for a triple threshold is
statistically significant with a p-value of 0.02.The p-value is least for the test of a
triple threshold effect, therefore we conclude that there exists a triple threshold effect
11

between the debt ratio and operating efficiency.

Table 2 Tests for threshold effects
Test for single threshold
19.446401

F1
P-Value

0.03
(10%, 5%, 1% critical value)

(11.34,14.4,20.68)

Test for double threshold

F2

10.198337

P-Value

0.06

(10%, 5%, 1% critical value)

(9.14, 10.78, 15.28)

Test for triple threshold

F3

10.127469

P-Value

0.02

(10%, 5%, 1% critical value)

(6.9,8.63,10.96)

denotes significance at the 10%,5%,1% level respectively.

When there exists a triple threshold effect between the debt ratio and operating
efficiency, all observations are split into four regimes. Table 2 represents the threshold
estimates and Table 4 represents the regression slope estimates, conventional OLS
standard errors, and White-corrected standard errors for four regimes.
Table 3 Threshold estimate
Estimate

0.21654495

0.35053867

0.51936048

Table 4 Regression estimates: triple threshold model

1
2

Coefficient
estimate

OLS Se

White Se

0.88791656
0.51746028

0.14039845(6.3)
0.068485810(7.65)

0.32423321(2.75)
0.20014241(2.59)

12

 3
4
1.

0.34217229
0.25437742

0.040440677(8.5)
0.016846981(15.1)

0.13830329(2.474)
0.10603472(2.399)

denotes at the significance of 1% level.

d it 1 , 2 is the coefficient estimate for 1 < d it 2 ,

3 is the coefficient estimate for 2< d it 3 and 4 is the coefficient estimate for d it > 3 .

2.

is the coefficient estimate for

3. The value inis tvalue.

Figure 1. Confidence interval construction in triple threshold model.

The estimated model from empirical research is as follows:

v it = i + 0.89 d it I ( d it 0.22 ) + 0.52 d it I (0.22 < d it 0.35) +

0.34 d it I (0.35 < d it 0.52) + 0.25d it I ( d it > 0.52 ) + it

, 2 and 3 split the observations into four regimes. In the first regime, where

the debt ratio is below 22%, the estimate of coefficient 1 is 0.89, which means,
when the debt ratio is increased by 1%, ROA will be increased by 0.89%. In the
second regime, where the debt ratio is above 22% but below 35%, the estimate of
coefficient 2 is 0.52, which implies, ROA will be increased by 0.52% by 1%
increase of debt ratio. In the third regime, where the debt ratio is above 35% but
below 52%, the estimate of coefficient 3 is 0.34, that is, when debt ratio is
increased by 1%, ROA will be increased by 0.34%. In the fourth regime, where debt
ratio is above 52%, the estimate of coefficient 4 is 0.25, that is, ROA will be
13

increased by 0.25% by 1% increase of debt ratio. According to table 4, the estimates

1 , 2 , 3 , 3 are all significant under the 1% level under the consideration of both
homogenous standard errors and heterogeneous standard errors.
From empirical research we find that the increase of debt ratio can improve the
operating efficiency of firms thus increase firm value. This finding is consistent with
the proposition MM (1963). On the other side, costs of financial distress increase as
the debt ratio increase gradually, which counteracts the positive effect of debt ratio to
operating efficiency. Our finding 4 < 3 < 2 < 1 is consistent with this view.
This result can also be explained by agency theory. On one hand, higher debt ratio
means less free cash flow for managers to manipulate, which can reduce agency cost
and increase firms operating efficiency. On the other hand, when the debt ratio is too
high which leads to financial distress, the conflicts between creditors and shareholders
will increase the agency costs and reduce the operating efficiency. Therefore, we
conclude that if financial managers use financial leverage deliberately, operating
efficiency will be improved and firm value will be increased.

5.Conclusions
The paper applies panel threshold regression model to examine whether financial
leverage affects operating efficiency thus affects firm value of Chinese listed firms.
We intend to test whether there exists an optimal debt ratio, which leads to the
asymmetric relationship between the debt ratio and operating efficiency.
Under the consideration of the special features of Chinese stock market, we choose
ROA as the indicator of the operating efficiency of the firms to test threshold effect
based on their debt to asset ratio. We find that there exists triple threshold effect and
the coefficients are all positive and significant for four regimes. Our results identify
that the operating efficiency is positively related to debt level, with very low debt
firms having a higher coefficient. but the positive effect decreases when debt ratio
increases. When debt ratio increases, the interest tax shield increases; however, on the
other side, costs of financial distress increase to counteract the positive effect of debt
ratio to firm value. As Chinese listed firms prefer equity financing and their debt to
asset ratios are relatively low, the benefits of interest tax shield are higher than costs
of financial distress. This result can also be explained by agency theory. On one hand,
higher debt ratio means less free cash flow for managers to manipulate, which can
reduce agency cost and increase firms operating efficiency. On the other hand, when
the debt ratio is too high which leads to financial distress, the conflicts between
14

creditors and shareholders will increase the agency costs and reduce the operating
efficiency.
Our results indicate that financial managers should use financial leverage wisely to
improve the operating efficiency in order to maximize the firm value and investors
should refer to the debt ratio to make investment decisions. As corporate bond market
has not been well developed in China and firms debt financing mainly depends on
commercial banks, our result is helpful for loan managers in banks.
References
(1)Bergman, Yaacov Z., and Jeffrey L. Callen, 1991, Opportunistic Underinvestment
in Debt Renegotiation and Capital Structure, Journal of Financial Economics, 29(1),
137-172
(2)Bradley, Michael, Gregg A. Jarrell, E. Han Kim and Wayne H. Mikkelson, 1984,
On the Existence of an Optimal Capital Structure: Theory and Evidence/Discussion,
The Journal of Finance, 39(3), 857-871
(3)Burgman, Todd A., 1996, An empirical examination of multinational corporate
capital structure, Journal of International Business Studies, 27(3), 553-471
(4)Castanias, Richard, 1983, Bankruptcy Risk and Optimal Capital Structure, The
Journal of Finance, 38(5), 1617-1636
(5)Chan, K.S., 1993, Consistency and Limiting Distribution of the Least Squares
Estimator of a Continuous Threshold Autoregressive Model, The Annals of Statistics,
21, 520-533
(6)Chirinko, Robert S. and R. Singha Anuja, 2000, Testing static tradeoff against
pecking order models of capital structure: A critical comment, Journal of Financial
Economics, 58(3), 417
(7)Davis, R. B., 1977, Hypothesis testing when a nuisance parameter is present only
under the alternative, Biometrika, 64, 247-254
(8)Davis, R. B., 1987, Hypothesis testing when a nuisance parameter is present only
under the alternative, Biometrika, 74, 33-43
(9)Erik Lie, 2002, Do firms undertake self-tender offers to optimize capital
structure? The Journal of Business, 75(4), 609-640
(10)Erwan, Morellec, 2001, Asset liquidity, capital structure, and secured debt,
Journal of Financial Economics, 61(2), 173
(11)Flath, David, and Charles R. Knoeber, 1980, Taxes, Failure Costs, and Optimal
Industry Capital Structure: An Empirical Test, The Journal of Finance, 35(1), 99
(12)Hansen, B. E., 1996, Inference when a nuisance parameter is not identified under
the null hypothesis, Econometrica, 64, 413-430
15

(13)Hansen, B. E., 1999, Threshold effects in non-dynamic panels: Estimation,

testing and inference, Journal of Econometrics, 93, 345-368.
(14)Hansen, B. E., 1986, Sample splitting and threshold estimation, Econometrica,
2000, 68, 575-603
(15)Jensen, Michael C., Agency Costs of Free Cash Flow, Corporate Finance, and
Takeovers, The American Economic Review, 76, 323-329
(16)Jensen, M.C., and W. H. Meckling, Theory of the firm : Managerial Behavior,
agency cost and ownership structure, Journal of Financial Economics, 1976, 3,
305-360
(17)Kim, Wi Saeng, and Eric H. Sorensen, 1986, Evidence on the Impact of the
Agency Costs of Debt on Corporate Debt Policy, Journal of Financial and
Quantitative Analysis
(18)Lakshmi, Shyam-Sunder, and Stewart C. Myers, 1999, Testing static tradeoff
against pecking order models of capital structure, Journal of Financial Economics,
51(2), 219-245
(19)Miller, Merton H., 1977, Dedt and Taxes, Journal of Finance, 32, 261-275
(20)Modigliani, F. and Merton H. Miller, 1958, The cost of capital, corporate finance,
and the theory of investment, American Economic Review, 48, 261-297
(21)Modigliani, F. and Merton H. Miller, 1963, Corporate income taxes and the cost
of capital: A correction, American Economics Review, 53, 433-443
(22)Myers, Stewart C., and Nicholas S. Majluf, 1984, Corporate Financing and
Investment Decisions When Firms Have Information That Investors Do Not Have,
Journal of Financial Economics, 13(2), 187-222
(23)Philosophov, L.V. and V.L. Philosophov, 1999, Optimization of corporate capital
structure: A probabilistic Bayesian approach, International Review of Financial
Analysis
(24)Ross, Stephen A., 1977, The Determination of Financial Structure-The Incentive
Signaling Approach, Bell Journal of Economics, 8(1), 1-23
(25)Titman, Sheridan, and Roberto Wessels, 1988, The Determinants of Capital
Structure Choice, The Journal of Finance, 43(1), 1-21
(26)Tong, H., On a Threshold Model, in C.H. Chen (ed.), Pattern Recognition and
Signal PROAessing, Amsterdam: Sijthoff & Noordhoff, 1978, 101-141.
(27) Chien-Chung Nieh, and Wen-qian Liu, Nonlinear relationship between capital
structure and operating efficiency of electron industriesPanel threshold effect
analysis
(28) Hui Wang, Debt financing, Corporate Governance and Firm Value, Chinese
Edition of Economic Review, 2003.8

16