Macroeconomic Conditions and Capital Structure

over the Business Cycle: Further Evidence in the

Context of Taiwan
Hsien-Hung H. Yeh and Eduardo Roca
ABSTRACT: The authors examine the effect of macroeconomic conditions on the capital
structure of firms in the petrochemical, textile, and electronics industries of Taiwan during
the period 19832007, which covers six and a half business cycles. Controlling for the
effects of economic growth, industry type, and firm-specific factors, we find that macroeconomic conditions have a significant effect on debt ratios during the sample period. Debt
ratios of firms in these industries are procyclical during the period before the 199798 Asian
financial crisis, but countercyclical during the period after the Asian financial crisis.
KEY WORDS: capital structure, macroeconomic conditions, Taiwan.

Over the past several decades, there have been numerous studies on capital structure.
Focusing mainly on the effects of factors at the firm and industry levels, these studies
have documented common determinants of capital structure (Harris and Raviv 1991).
These studies have identified a number of firm-specific factors that affect capital structure, namely, firm size, growth opportunities, profitability, earnings volatility, nondebt
tax shields, and asset tangibility. However, some of them, for example, profitability and
growth opportunities, vary with macroeconomic conditions over the business cycle. This
suggests, therefore, that capital structure may vary with macroeconomic conditions.
Jensens free cash flow theory (1986) contended that corporate debt financing enables
managers to effectively bound their promise to pay out future cash flows. Thus, debt
can reduce the agency cost of free cash flow because debt decreases the cash flow available for spending at the discretion of managers. Jensen asserted that the threat caused
by failure to make debt service payments can serve as an effective motivating force to
make organizations more efficient. Thus, based on Jensens free cash flow theory, firms
would finance more debt during economic expansion due to the increase in cash flow,
but less debt during economic contraction due to the decrease in cash flow. This implies
that capital structure is procyclical over the business cycle.
On the contrary, Miller (1977) found variation in the debt ratios of nonfinancial
companies over the business cycles between the 1920s and the 1960s. He concluded
that the cyclical movements of the economy result in lower debt ratios during economic
expansion. Consistent with the findings of Miller (1977), Korajczyk and Levy (2003),
Hsien-Hung H. Yeh (hhyeh@mail.npust.edu.tw) is an associate professor in the Department of
Business Administration, National Pingtung University of Science and Technology, Pingtung,
Taiwan. Eduardo Roca (e.roca@griffith.edu.au) is a professor in the Department of Accounting,
Finance, and Economics, Griffith University, Queensland, Australia. The authors are grateful for
the valuable comments and suggestions received from the anonymous referees and the editor.
They also thank Michael Drew, Kurt Dew, and Graham Bornholt of Griffith University for helpful
comments. In addition, Hsien-Hung H. Yeh thanks the National Pingtung University of Science
and Technology, Taiwan, and the Griffith Business School of Griffith University, Australia, for
financial assistance.
Emerging Markets Finance & Trade / SeptemberOctober 2012, Vol. 48, Supplement 3, pp. 141156.
2012 M.E. Sharpe, Inc. All rights reserved. Permissions: www.copyright.com
ISSN 1540496X (print)/ISSN 15580938 (online)
DOI: 10.2753/REE1540-496X4805S309

142 Emerging Markets Finance & Trade

with financial constraints taken into account, examined the effect of macroeconomic
conditions on capital structure decisions and found a countercyclical or negative effect
on corporate leverage ratios for financially unconstrained firms. Hackbarth et al. (2006)
analyzed capital structure with credit risk under consideration in their model and contended
that default thresholds are countercyclical and, thus, stated that market leverage ratios
should be countercyclical. Levy and Hennessy (2007) analyzed corporate financing over
business cycles in their general equilibrium model and contended that capital structure
is countercyclical for the less constrained firms. It appears therefore that, theoretically,
there is no agreement as to the effect of macroeconomic conditions on capital structure.
Hence, in this paper, we gather further evidence on the effect of macroeconomic conditions on capital structure.
Further, Glen and Singh (2004) found that capital structure in emerging countries
is different from that observed in developed countries. However, most evidence on
the effect of macroeconomic conditions on capital structure comes from developed
countries rather than developing countries or emerging markets (C +rnigoj and Mramor
2009; Hazak 2009; Nieh et al. 2008). For example, in the investigation of the target
capital structure of listed Taiwanese electronics firms, Nieh et al. (2008) found a single
threshold effect of debt ratio on return on equity that is used as a proxy for firm value.
Nieh et al. also found that the optimal debt ratio range for listed Taiwanese electronics
firms should not be over 51.57 percent or below 12.37 percent. Further, in a Slovenian
empirical study, C +rnigoj and Mramor (2009) concluded that, throughout the period
1999 to 2006, the financial behavior of Slovenian firms most probably remained different from those in mature market economies. In addition, they found that the financial
behavior of Slovenian firms was changing during the transition. Still further, Hazak
(2009) analyzed corporate capital structure and dividend decisions under distributed
profit taxation in Estonia. This study found that the distributed profit taxation system
led Estonian firms to pay fewer dividends and retain more earnings and, in addition,
the importance of external financing in total capital of the sample firms had decreased.
These previous studies, however, did not take into account the effect of macroeconomic
conditions on capital structure.
This paper, therefore, addresses this important gap in the literature through an
examination of the effect of macroeconomic conditions on capital structure in the
context of Taiwan with a focus on the textile, petrochemical, and electronics industries.
Taiwan is a country that has had a very successful experience of economic transition
from being an emerging country to becoming a developed one. The country adopted a
selective industry policy from the 1950s to the 1990s, which is recognized as having
played an important role in its economic development (Chu 1994, 2003). The textile,
petrochemical, and electronic industries served as the cornerstones of this industrial
policy. As far as we know, no study examining the impact of macroeconomic conditions on these industries in the context of Taiwan has been conducted. In this paper,
we investigate the effect of macroeconomic conditions on capital structure over the
business cycles before and after the 199798 Asian financial crisis in relation to the
textile, petrochemical, and electronics industries of Taiwan. In particular, we seek to
answer the following questions:

1. Did macroeconomic conditions affect capital structure for periods before and
after the 199798 Asian financial crisis?
2. If so, did the effect of macroeconomic conditions on capital structure vary
across periods before and after the Asian financial crisis?

SeptemberOctober 2012 Supplement 143

Literature Review
After the basic work of Modigliani and Miller (1958), extensive studies have been
conducted on capital structure decisions, as summarized by Harris and Raviv (1991).
These previous studies have addressed the issue at the firm and industry levels and have
found common, firm-specific determinants of capital structure, which include firm size,
growth opportunities, profitability, nondebt tax shields, and asset tangibility. Economic
output falls during the period of recession but increases during the period of expansion,
particularly at an economic trough and peak. Some firm-specific determinants of capital
structure, for example, profitability and growth opportunities, may also vary with the state
of the economy over the business cycle, as there are more future growth opportunities
available to firms at an economic trough than at an economic peak. The link between
macroeconomic conditions, firm-level factors, and capital structure therefore suggests
that capital structure would be related to macroeconomic conditions.
Jensens free cash flow theory (1986) asserted that debt can be used to motivate managers to make their organizations more efficient in the presence of large free cash flows.
Jensen contended that debt creation enables managers to effectively bond their promise
to pay out future cash flows, and thus debt can be an effective substitute for dividend.
Debt reduces the agency cost of free cash flow because it reduces the cash flow available
for spending at the discretion of managers. The threat caused by failure to make debt
service payments can serve as an effective motivating force to improve organizations
efficiency. Hence, to reduce the agency problem between management and shareholders
in order to achieve organizational efficiency, firms would finance with more debt during
economic expansion due to the increase in cash flow, but with less debt during economic
contraction due to the decrease in cash flow. This implies that capital structure would be
procyclical over the business cycle.
However, Miller (1977) reported that debt ratios of the typical nonfinancial companies varied with the business cycles between 1920 and 1960 and, in addition, debt ratios
tended to fall during economic expansions. Korajczyk and Levy (2003) examined the
impact of macroeconomic conditions on capital structure. They split their sample into
two subsamples, financially constrained and financially unconstrained, which allowed
them to test whether the trade-off theory and the pecking order theory can explain the
effects of financial constraints and macroeconomic conditions on capital structure decisions. Korajczyk and Levy found that corporate leverage is countercyclical for financially
unconstrained firms. Hackbarth et al. (2006) developed a contingency-claims model and
analyzed credit risk and capital structure. They argued that shareholders value-maximization default policy is characterized by a different threshold for each state and, in addition,
default thresholds are countercyclical. Their model predicted that market leverage should
be countercyclical. Further, Levy and Hennessy (2007) developed a general equilibrium
model for corporate financing over the business cycle. Levy and Hennessy argued that
managers would hold a proportion of their firms equity, that is, managerial equity shares,
in order to avoid agency conflicts. Because of the increases in managerial wealth and risk
sharing, firms finance with less debt during expansions than during contractions. Based
on their simulations, Levy and Hennessy found a countercyclical variation in leverage
for less constrained firms. Their finding of a negative effect of macroeconomic conditions
on capital structure is consistent with Korajczyk and Levy (2003).
Based on the above discussion, it seems therefore that, theoretically, there is no agreement as to the effect of macroeconomic conditions on capital structure. Furthermore,
Glen and Singh (2004) found that the capital structure of firms in developed countries

144 Emerging Markets Finance & Trade

is different from that in emerging markets. However, most previous studies of the effect
of macroeconomic conditions on capital structure were conducted within the context of
developed countries rather than developing countries or emerging markets. Thus, this
paper addresses this gap by providing further evidence on this issue in the context of
Taiwan, which became newly industrialized at the beginning of this century.
Methods and Data
The Model for the Determination of Capital Structure
As suggested by previous studies (Chen 2004; Chu et al. 1992; Feidakis and Rovolis
2007; Hackbarth et al. 2006; Harris and Raviv 1991; Korajczyk and Levy 2003; Kuo and
Wang 2005), firm characteristics, industry type, economic growth and macroeconomic
conditions are related to capital structure. Moreover, the findings in previous studies on
the adjustment of capital structure (Flannery and Rangan 2006; Ozkan 2001; Qian et
al. 2009) show that debt ratios are positively related to their lagged values in the capital
structure adjustment process, although there is ongoing debate as to whether firms adjust
toward the target (Chang and Dasgupta 2009; Hovakimian and Li 2011). Therefore, we
assume that firm characteristics, industry type, economic growth and macroeconomic
conditions, and the previous level of capital structure linearly determine the capital
structure of firms.
Taking into account macroeconomic conditions (EC) and economic growth (EG),
industry type (IND), firm-specific factors, including firm size (SIZE), growth opportunities (GROWTH), profitability (PROFIT), nondebt tax shields (NDTS), asset tangibility
(ASSET), and the previous level of capital structure (CSt1), the model for the determination of capital structure (CS) can be written as follows:

CSit = b0 + bECECit + bEGEGit + bINDINDit + b1SIZEit + b2GROWTHit

+ b3PROFITit + b4NDTSit + b5ASSETit + b6CSit1 + eit.

(1)

Based on Equation (1), the coefficient on the binary dummy variable EC, which represents macroeconomic conditions, is expected to be negative based on the finding of
previous studies that macroeconomic conditions have a countercyclical effect on capital
structure (Hackbarth et al. 2006; Korajczyk and Levy 2003; Levy and Hennessy 2007).
To further investigate whether the effect of macroeconomic conditions on capital
structure varies across periods, we expand Equation(1) with an additional specification
by including the dummy proxy for periods before and after the 199798 Asian financial
crisis (AFC) and the product of the proxies for macroeconomic conditions and the Asian
financial crisis, (ECAFC), as follows:

CSit = b0 + bECECit + bAFCAFCit + bECAFCECAFCit + + bEGEGit

+ bINDINDit + b1SIZEit + b2GROWTHit + b3PROFITit
+ b4NDTSit + b5ASSETit + b6CSit1 + eit.

(2)

Variables and Measures

All financial ratios used in the study are calculated at book value of the collected financial
data. As suggested by most of the previous studies (Chen 2004; Harris and Raviv 1991;
Qian et al. 2009), the ratio of total debt to total assets, that is, debt ratio (DR), is used as

SeptemberOctober 2012 Supplement 145

the proxy for the capital structure of firms. To examine the effect of economic conditions
on capital structure, the years during the periods from economic troughs to economic peaks
and during the periods from economic peaks to economic troughs in the sample period of
1983 to 2007 are selected to represent economic expansion and contraction, respectively.
The binary dummy variable EC with a value of 1 and 0 for economic expansion and
contraction, respectively, is used as the proxy for the shift in macroeconomic conditions.
The classification of economic expansion and contraction is based on the reference dates
of business cycles in the Taiwan Business Indicators that are officially published by the
Council for Economic Planning and Development of Taiwan (CEPD 2009).
Moreover, the binary dummy AFC, with values of 0 and 1, is used to denote the
period before and after the 199798 Asian financial crisis, respectively, while the annual growth rate of gross domestic product (gGDP) is used as the proxy for economic
growth (Feidakis and Rovolis 2007). Bradley et al. (1984) found a significant difference
in firm leverage ratios. Among the twenty-five selected industries in their sample, the
mean debt ratios for the electronics, petroleum refining, and textile industries were 16
percent, 24percent, and 33 percent, respectively. Their finding suggests that firms in the
electronics industry tend to finance with less debt than those in the petrochemical and
textile industries. In our study, if they have a value of 1, the binary dummies IND13 and
IND14 denote the petrochemical and textile industries, respectively, and the electronics
industry when their value is 0. The natural logarithm of total assets (lnTA) is used to
measure firm size (Chen 2004; Kuo and Wang 2005) and the annual growth rate of total
assets (gTA) proxies for growth opportunities. Profitability is represented by the ratio of
operating income to total assets (OITA) (Titman and Wessels 1988). The ratio of total
depreciation to total assets (DEPTA) is used as the proxy for nondebt tax shields (Chen
2004; Chu et al. 1992; Kim and Sorensen 1986; Qian et al. 2009; Titman and Wessels
1988; Wald 1999; Wiwattanakantang 1999). The ratio of inventory plus net fixed assets to
total assets (INVFATA) is used to measure collateral value or tangibility of asset structure
(Chen 2004; Chu et al. 1992; Downs 1993; Titman and Wessels 1988; Wald 1999).
Estimation
Because of the presence of the one-period lagged dependent variable as a regressor in
our dynamic panel data model, the lagged dependent variable in the equation is correlated with the error term. The ordinary least squares (OLS) estimator is therefore biased
and inconsistent. Further, the instrumental variable estimation method can lead to the
estimator being consistent but not necessarily efficient in the dynamic panel model. This
is partly because the differenced structure in the disturbances is not considered in the
instrumental variable estimation method (Baltagi 2008). With instrumental variables taken
into account in the specification of a dynamic panel model, the generalized method of
moments (GMM) estimation can lead to consistent and efficient estimates of parameters
(Baltagi 2008).
However, the standard GMM robust two-step estimator of the variancecovariance
matrix (VCE) can be seriously biased. Windmeijer (2005) found from his Monte Carlo
studies that estimated standard errors of the efficient two-step generalized method of
moments (GMM) estimator can be severely biased downward in small samples. He
derived a bias-corrected robust estimator, that is, a WC robust estimator, for two-step
VCEs from GMM estimators, which are robust to heteroskedasticity in the errors. Thus,
in this study we use the two-step dynamic panel data GMM (DPD-GMM) estimation

146 Emerging Markets Finance & Trade

that can be implemented in the linear dynamic panel data estimation (i.e., XTDPD) of
Stata. Instrumental variables that are used in our DPD-GMM estimation include macroeconomic conditions, economic growth, industry type, firm-specific variables, and
the lagged dependent variable. More detailed information in relation to the DPD-GMM
estimation can be found in the Stata users guide (Stata 2009).
Sample and Data
The sample includes the listed firms in the petrochemical, textile, and electronics industries of Taiwan that have complete financial data for six and one-half business cycles
during the period 19832007. Annual financial data of firms in the selected industries
are collected from the database of the Taiwan Economic Journal. However, firms that
experienced financial distress or trade suspension are excluded. The sample consists of
137 listed firms in the selected industries over the research period. In the sample, there
are sixteen firms from the petrochemical, forty-eight from the textile, and seventy-three
from the electronics industries. In total, there are 2,718 observations in the sample over
the period 19832007, of which 1,485 observations pertain to the period 198398, which
covers four business cycles, and 1,233 observations relate to the period 19992007, which
encompasses two and a half business cycles after the 199798 Asian financial crisis.
Unbalanced panel data are used in the study to examine the effect of macroeconomic
conditions on capital structure across periods before and after the Asian financial crisis
of 199798.
Descriptive Statistics and Correlation Matrix
Table 1 provides summary descriptive statistics for the variables in the study. As can be
seen in PanelA of Table 1, the average debt ratio of firms in the sample industries for
the full sample period 19832007 is about 40.85 percent. For the subperiods, 198398
(pre-financial crisis) and 19992007 (post-financial crisis), the average debt ratio for the
firms is 42.59 percent and 38.76 percent, respectively, as can be seen in Panels B and C
of the table. This shows that the average debt ratio for firms in the three industries seems
to decline after the financial crisis.
The correlation matrix among debt ratios, economic variables, and firm-specific
variables is presented in Table 2. As shown in the table, the binary dummy proxy for
macroeconomic conditions (EC) is not significantly related to the debt ratios (DR) of
firms in the selected industries. This indicates that macroeconomic conditions seem to be
unrelated to the debt ratios of firms in the selected industries during the period 19832007.
The proxy for economic growth (gGDP) is significantly and positively related to the
debt ratios. The dummy IND13, the proxy for the petrochemical industry, is statistically
significant and negatively related to the debt ratios; however, the dummy IND14, which
is a proxy for the textile industry, is not significantly related to the debt ratios. These
results suggest that the debt ratios of firms may have varied across these three industries
during the period 19832007. Moreover, all of the proxies for firm characteristics are
significantly related to the debt ratios and the one-period lagged debt ratios (DRlag) are
positively related to the debt ratios, which is in line with the finding of previous studies
on the adjustment of capital structure. This confirms that it is appropriate for us to include
the one-period lagged debt ratios in the model.

SeptemberOctober 2012 Supplement 147

Table 1. Summary descriptive statistics

Variable

Mean

Panel A: Full sample for the period 19832007

DR
2,718
0.4085
EC
2,718
0.7553
gGDP
2,718
0.0582
IND13
2,718
0.1314
IND14
2,718
0.3668
lnTA
2,718
22.3214
gTA
2,718
0.1922
OITA
2,718
0.0543
DEPTA
2,718
0.0342
INVFATA
2,718
0.4669
DRlag
2,718
0.4193
Panel B: Subsample for the period 19831998
DR
1,485
0.4259
EC
1,485
0.7367
gGDP
1,485
0.0684
IND13
1,485
0.1434
IND14
1,485
0.3805
lnTA
1,485
21.7642
gTA
1,485
0.2861
OITA
1,485
0.0766
DEPTA
1,485
0.0368
INVFATA
1,485
0.5413
DRlag
1,485
0.4454
Panel C: Subsample for the period 19992007
DR
1,233
0.3876
EC
1,233
0.7778
gGDP
1,233
0.0460
IND13
1,233
0.1168
IND14
1,233
0.3504
lnTA
1,233
22.9926
gTA
1,233
0.0791
OITA
1,233
0.0275
DEPTA
1,233
0.0310
INVFATA
1,233
0.3772
DRlag
1,233
0.3879

Standard
deviation

Minimum

Maximum

0.16335
0.42997
0.02385
0.33784
0.48202
1.41975
0.39614
0.07745
0.02603
0.20434
0.16825

0.01545
0
0.01651
0
0
17.66759
0.54130
0.28382
0
0.00011
0.01396

0.96655
1.00000
0.11004
1.00000
1.00000
27.15450
7.12272
0.59860
0.22850
0.90002
0.98689

0.16938
0.44057
0.01903
0.35063
0.48567
1.21696
0.43172
0.08192
0.02374
0.17477
0.17664

0.06042
0
0.03466
0
0
17.66759
0.45089
0.28382
0
0.06397
0.01396

0.96655
1.00000
0.11004
1.00000
1.00000
25.86359
5.52098
0.59860
0.22850
0.90002
0.98689

0.15324
0.41591
0.02330
0.32130
0.47728
1.35508
0.31331
0.06188
0.02824
0.20139
0.15179

0.01545
0
0.01651
0
0
20.47627
0.54130
0.26956
0.00031
0.00011
0.01545

0.90210
1.00000
0.06190
1.00000
1.00000
27.15450
7.12272
0.29435
0.15586
0.87153
0.90210

Notes: DR = total debts/total assets; EC = 0 and 1 for economic contraction and expansion,
respectively; gGDP = annual growth rate of real gross domestic product; IND13 = 1 for petrochemical
industry; IND14 = 1 for textile industry; lnTA = natural logarithm of total assets; gTA = annual growth
rate of total assets; OITA = net operating income/total assets; DEPTA = depreciation/total assets;
INVFATA = inventory plus net fixed assets/total assets; DRlag = total debt ratio at the previous year.

Regression Results
The two-step dynamic panel data GMM regression results for the debt ratios of firms
in the petrochemical, textile, and electronics industries over the business cycles during
the subperiods 198396, 198398 and 19992007 are shown in Table 3. As shown in

1.00000
0.00475
0.8043
0.11894
< 0.0001
0.12248
< 0.0001
0.00645
0.7368
0.09643
< 0.0001
0.06419
0.0008
0.13995
< 0.0001
0.08735
< 0.0001
0.19004
< 0.0001
0.85479
< 0.0001

0.10721
< 0.0001
0.00088
0.9636
0.00343
0.8582
0.03332
0.0824
0.06718
0.0005
0.00607
0.7518
0.04210
0.0282
0.00113
0.9529
0.00710
0.7113

1.00000

EC

0.04102
0.0325
0.03419
0.0747
0.29338
< 0.0001
0.15480
< 0.0001
0.23938
< 0.0001
0.07084
0.0002
0.25078
< 0.0001
0.15507
< 0.0001

1.00000

gGDP

0.29597
< 0.0001
0.08464
< 0.0001
0.10946
< 0.0001
0.02249
0.2411
0.04433
0.0208
0.03291
0.0862
0.11972
< 0.0001

1.00000

IND13

0.23459
< 0.0001
0.18000
< 0.0001
0.21992
< 0.0001
0.12269
< 0.0001
0.47407
< 0.0001
0.00895
0.6409

1.00000

IND14

0.07084
0.0002
0.11810
< 0.0001
0.02366
0.2176
0.36619
< 0.0001
0.14421
< 0.0001

1.00000

lnTA

0.35014
< 0.0001
0.17150
< 0.0001
0.04332
0.0239
0.05630
0.0033

1.00000

gTA

0.02014
0.2940
0.01571
0.4131
0.00080
0.9668

1.00000

OITA

0.51253
< 0.0001
0.05977
0.0018

1.00000

DEPTA

0.18526
< 0.0001

1.00000

INVFATA

1.00000

DRlag

Notes: Sample size = 2,718. DR = total debts/total assets; EC = 0 and 1 for economic contraction and expansion, respectively; gGDP = annual growth rate of real gross
domestic product; IND13 = 1 for petrochemical industry; IND14 = 1 for textile industry; lnTA = natural logarithm of total assets; gTA = annual growth rate of total assets; OITA = net operating income/total assets; DEPTA = depreciation/total assets; INVFATA = inventory plus net fixed assets/total assets; DRlag = total debt ratio at the
previous year. Probability>|Rho| under H0: Rho=0 is shown under the correlation coefficients.

DRlag

INVFATA

DEPTA

OITA

gTA

lnTA

IND14

IND13

gGDP

DR
EC

DR

Table 2. Correlation matrix

148 Emerging Markets Finance & Trade

6.393***
1.006
1,507***

6.916***
0.645
1,396***

133.091
1.000

0.094
0.005
0.214
0.014
0.010
0.004
0.015
0.043
0.169
0.028
0.030

WC-robust
standard error

127.033
0.978

0.026
0.010*
0.448**
0.040***
0.052***
0.005
0.028*
0.354***
0.818***
0.187***
0.620***

Beta
coefficient

1,348

0.094
0.006
0.167
0.012
0.009
0.004
0.016
0.045
0.175
0.029
0.028

WC-robust
standard error

1,074

0.037
0.017***
0.427***
0.037***
0.051***
0.005
0.047***
0.369***
0.838***
0.174***
0.638***

Beta
coefficient

(B) 19831998

5.917***
0.376
608***

0.087
0.005
0.088
0.013
0.012
0.004
0.036
0.074
0.293
0.037
0.044

WC-robust
standard error

1,096
90.543
0.122

0.143*
0.013**
0.001
0.011
0.026**
0.011***
0.024
0.463***
0.655**
0.126***
0.702***

Beta
coefficient

(C) 19992007

Notes: EC = 0 and 1 for economic contraction and expansion, respectively; gGDP = annual growth rate of real gross domestic product; IND13 = 1 for petrochemical industry; IND14 = 1 for textile industry; lnTA = natural logarithm of total assets; gTA = annual growth rate of total assets; OITA=net operating
income/total assets; DEPTA = depreciation/total assets; INVFATA = inventory plus net fixed assets/total assets; DRlag = total debt ratio at the previous year.
The GMM estimates reported are all two-step estimates. Instrumental variables include EC, gGDP, IND13, IND14, lnTA, gTA, OITA, DEPTA, INVFATA, and
DRlag. The Wald test is used to test joint significance of the independent variables asymptotically distributed as c2(k) under the null hypothesis of no relationship. ***Significant at 1 percent; **significant at 5 percent; *significant at 10 percent.

Constant
EC
gGDP
IND13
IND14
lnTA
gTA
OITA
DEPTA
INVFATA
DRlag
Number of observations
Sargan test
J statistic
p-value
First-differenced residual
correlation
First-order
Second-order
Wald test: c2

Variable

(A) 19831996

Table 3. Dynamic panel-data GMM regression results for debt ratios for periods before and after the 19971998 Asian
financial crisis

SeptemberOctober 2012 Supplement 149

150 Emerging Markets Finance & Trade

the table, the ArellanoBond test (1991) performed with the XTDPD command of Stata
suggests that it is appropriate for us to use the DPD-GMM estimation in this study because of the absence of second-order residual correlation. In addition, the Sargan test
of overidentifying restrictions for instrumental variables in the DPD-GMM estimation
indicates that overidentifying restrictions are valid. Moreover, the Wald test for testing the
joint significance of independent variables suggests that the debt ratios are significantly
influenced by one or more independent variables in the dynamic panel data model.
Further, as can be seen in column A of Table 3, controlling for firm-specific and industry
effect, macroeconomic conditions (EC) and economic growth (gGDP) are significantly
and positively related to debt ratios (DR) of firms in these three selected industries for
the period 198396 (before the 199798 Asian financial crisis). Similar results are found
for the period 198398, as shown in column B of the table. These findings suggest that,
controlling for the effect of economic growth as well as firm-specific and industry effect,
macroeconomic conditions have a positive effect on debt ratios of firms in the sample
industries before the 199798 Asian financial crisis. However, macroeconomic conditions
are significantly and negatively related to debt ratios of firms in the sample industries for
the period 19992007. As a whole, the above findings suggest that the debt ratios of firms
in the petrochemical, textile, and electronics industries of Taiwan varied across periods
before and after the 199798 Asian financial crisis.
Moreover, as shown in column C of Table 3, the proxy for economic growth (gGDP)
is not significantly related to the debt ratios of firms in the sample industries for the
period 19992007 (after the 199798 Asian financial crisis). It is possible that the debt
ratios of firms in these industries may be affected by the economic development and/or
financial development of Taiwan (Akimov and Dollery 2008; Boyd and Smith 1996). In
particular, Boyd and Smith (1996) concluded that the aggregate ratio of debt to equity
will generally decline as an economy develops. Future research may provide evidence
of the effect of economic and financial development on capital structure.
In addition, as shown in columns A and B of Table 3, the results show that the debt
ratios in the petrochemical and textile industries are significantly lower than those in the
electronics industry during the period before the Asian financial crisis. This finding is not
in line with that of Bradley et al. (1984). It is likely that firms in the electronics industry
finance more debt than those in the petrochemical and textile industries because of Taiwans industrial policy in favor of the selective industry during the pre-financial crisis
period. However, during the period after the Asian financial crisis, the debt ratios in the
electronics industry are significantly higher than those in the textile industry only. This is
most probably because of the change in the government industrial policy, as mentioned
earlier in the introduction. Future research may provide further evidence of the effect of
industrial policy on capital structure decisions. Finally, as can be seen in the table, the
results regarding the effect of firm characteristics on debt ratios are consistent with the
findings of most previous studies.
Further regression results based on panel least square estimation are reported in Tables4
and5. As can be seen in Table 4, for regression results that do not include the lagged debt
ratios in the model, the DurbinWatson dstatistic is far from 2, which suggests that the
estimator is biased and inconsistent due to serious residual autocorrelation.
However, as shown in the regression results presented in Table 5, when the lagged
debt ratios are included in the model, the explanatory power is greater than when lagged
debt ratios are excluded. However, the Lagrange multiplier (LM) test disproves the
null hypothesis of no residual autocorrelation. This suggests that the panel least square

0.356***
0.016
1.185***
0.100***
0.089***
0.004
0.039***
0.404***
1.909***
0.375***
1,211
0.220
0.399

Constant
EC
gGDP
IND13
IND14
lnTA
gTA
OITA
DEPTA
INVFATA
Sample size
Adj. R2
DurbinWatson d-statistic

0.095
0.011
0.280
0.014
0.011
0.004
0.011
0.057
0.209
0.033

Standard
error
0.291***
0.013
1.508***
0.103***
0.097***
0.001
0.025**
0.427***
1.934***
0.369***
1,485
0.227
0.451

Beta
coefficient
0.084
0.009
0.217
0.012
0.010
0.003
0.010
0.051
0.192
0.029

Standard
error

(B) Period: 19831998

0.014
0.016
0.169
0.038***
0.045***
0.018***
0.019
0.695***
1.372***
0.247***
1,233
0.126
0.273

Beta
coefficient

0.083
0.010
0.186
0.014
0.012
0.003
0.014
0.073
0.192
0.032

Standard
error

(C) Period: 19992007

Notes: EC = 0 and 1 for economic contraction and expansion, respectively; gGDP = annual growth rate of real gross domestic product; IND13 = 1 for petrochemical industry; IND14 = 1 for textile industry; lnTA = natural logarithm of total assets; gTA = annual growth rate of total assets; OITA = net operating
income/total assets; DEPTA = depreciation/total assets; INVFATA = inventory plus net fixed assets/total assets. ***Significant at 1 percent; **significant at
5percent; *significant at 10 percent.

Beta
coefficient

Variable

(A) Period: 19831996

Table 4. Panel least square results for debt ratios during the periods before and after the 19971998 Asian financial crisis
without lagged debt ratio in the model

SeptemberOctober 2012 Supplement 151

0.016
0.020***
0.264*
0.025***
0.038***
0.003
0.032***
0.385***
0.648***
0.117***
0.765***

Beta
coefficient

1,211
0.768
1.880
8.18***

0.052
0.006
0.154
0.008
0.006
0.002
0.006
0.031
0.117
0.018
0.014

Standard
error
0.028
0.014***
0.318**
0.023***
0.039***
0.004*
0.024***
0.385***
0.648***
0.125***
0.757***

Beta
coefficient

1,485
0.750
1.925
6.32**

0.048
0.005
0.125
0.007
0.006
0.002
0.006
0.029
0.112
0.017
0.014

Standard
error

(B) Period: 19831998

0.072*
0.014***
0.051
0.009
0.014**
0.006***
0.020***
0.323***
0.410***
0.093***
0.849***

Beta
coefficient

1,233
0.786
1.904
5.13**

0.041
0.005
0.092
0.007
0.006
0.002
0.007
0.036
0.097
0.016
0.014

Standard
error

(C) Period: 19992007

Notes: EC = 0 and 1 for economic contraction and expansion, respectively; gGDP = annual growth rate of real gross domestic product; IND13 = 1 for petrochemical industry; IND14 = 1 for textile industry; lnTA = natural logarithm of total assets; gTA = annual growth rate of total assets; OITA = net operating
income/total assets; DEPTA = depreciation/total assets; INVFATA = inventory plus net fixed assets/total assets; DRlag = total debt ratio at the previous year.
***Significant at 1 percent; **significant at 5percent; *significant at 10 percent.

Constant
EC
gGDP
IND13
IND14
lnTA
gTA
OITA
DEPTA
INVFATA
DRlag
Sample size
Adj. R2
DurbinWatson d-statistic
LM test: c2

Variable

(A) Period: 19831996

Table 5. Panel least square results for debt ratios during periods before and after the 19971998 Asian financial crisis
with lagged debt ratio included in the model

152 Emerging Markets Finance & Trade

SeptemberOctober 2012 Supplement 153

estimation for the model with one-period lagged debt ratios cannot lead to an efficient
and consistent estimator because of residual autocorrelation.
To provide further evidence as to whether the effect of macroeconomic conditions on
debt ratios varies across periods before and after the 199798 Asian financial crisis, we
expand our model by including a dummy proxy for the 199798 Asian financial crisis
(AFC), as well as the product of the proxies for macroeconomic conditions and the Asian
financial crisis (ECAFC). To check robustness, a DPD-GMM regression was performed.
The results for debt ratios at different periods are reported in Table 6.
As can be seen in Table 6, the ArellanoBond test for zero autocorrelation in firstdifferenced errors performed with the XTDPD command of Stata suggests that it is
appropriate for us to use DPD-GMM estimation in this study. The Sargan test of over
identifying restrictions for instrumental variables in the DPD-GMM estimation indicates
that overidentifying restrictions are valid. The Wald test for testing joint significance of
independent variables suggests that the debt ratios are significantly influenced by one
or more independent variables in the model for different periods of time. Moreover,
as shown in Table 6, debt ratios in these three selected industries are significantly and
positively related to the proxy for macroeconomic conditions (EC) over the business
cycles for different periods of time, that is, 19832007, 19852007, and 19892007.
This implies that macroeconomic conditions have a positive, procyclical effect on debt
ratios of firms in these three industries. The dummy proxy for the Asian financial crisis
(AFC) is positively and statistically related to the debt ratios at different periods of time.
This suggests that debt ratios in the selected industries increased after the 199798
Asian financial crisis. However, the proxy for the interaction between macroeconomic
conditions and the Asian financial crisis (ECAFC) has a significant, negative effect on
the adjustment of debt ratios at different periods. As a whole, the results illustrate that
the effect of macroeconomic conditions on debt ratios in these selected industries varied
across the periods before and after the 199798 Asian financial crisis. This finding is
consistent with those shown in Table 3.
Conclusion
Some recent studies on the effect of macroeconomic conditions on capital structure suggested that capital structure is countercyclical over the business cycle. However, Jensens
free cash flow theory contended that debt can reduce the agency cost of free cash flow
because debt reduces the cash flow available for spending at the discretion of managers
and debt financing can thus serve as an effective motivating force to make organizations
more efficient. Based on Jensens free cash flow theory, firms would finance with more
debt during economic expansion and less debt during economic contraction as a result
of the change in cash flow. This implies that capital structure would be procyclical over
the business cycle. Hence, it appears that there is no agreement as regards the effect of
macroeconomic conditions on capital structure. In this study, we provided fresh additional
evidence on this issue. We examined the effect of macroeconomic conditions on the debt
ratios of firms in the petrochemical, textile, and electronics industries of Taiwan over the
period of 19832007, which covers six and a half business cycles.
The results showed that the debt ratios of firms in these industries are significantly
influenced by macroeconomic conditions. We found that the debt ratios for firms are
procyclical over the business cycles during the period of 19831996, that is, before the
199798 Asian financial crisis, which supports Jensens free cash flow theory. However,

8.058***
0.248
1,642***

7.958***
0.117
2,112***

134.267
1.000

0.065
0.005
0.007
0.006
0.093
0.010
0.010
0.003
0.018
0.038
0.178
0.027
0.030

WC-robust
standard error

132.647
1.000

0.020
0.014***
0.027***
0.024***
0.111
0.029***
0.043***
0.003
0.023
0.378***
0.800***
0.163***
0.653***

Beta
coefficient

2,535

0.071
0.005
0.007
0.005
0.084
0.008
0.010
0.003
0.172
0.037
0.167
0.024
0.026

WC-robust
standard error

2,581

0.014
0.011**
0.023***
0.020***
0.108
0.028***
0.043***
0.004
0.021
0.383***
0.797***
0.166***
0.662***

Beta
coefficient

(B) 19852007

7.830***
0.129
1,380***

0.063
0.006
0.007
0.006
0.099
0.009
0.010
0.003
0.016
0.041
0.191
0.027
0.032

WC-robust
standard error

2,349
129.821
1.000

0.008
0.012**
0.023***
0.022***
0.065
0.032***
0.046***
0.005**
0.023
0.408***
0.803***
0.166***
0.631***

Beta
coefficient

(C) 19892007

Notes: EC = 0 and 1 for economic contraction and expansion, respectively; AFC = 0 and 1 for periods before and after the Asian financial crisis, respectively;
ECAFC= interaction between macroeconomic conditions and the financial crisis; gGDP = annual growth rate of real gross domestic product; IND13=1 for
petrochemical industry; IND14 =1 for textile industry; lnTA = natural logarithm of total assets; gTA = annual growth rate of total assets; OITA = net operating
income/total assets; DEPTA = depreciation/total assets; INVFATA = inventory plus net fixed assets/total assets; DRlag = total debt ratio at the previous year.
The GMM estimates reported are all two-step estimates. Instrumental variables include EC, gGDP, IND13, IND14, lnTA, gTA, OITA, DEPTA, INVFATA, and
DRlag. The Wald test is used to test joint significance of the independent variables asymptotically distributed as c2(k) under the null hypothesis of no relationship. ***Significant at 1 percent; **significant at 5 percent; *significant at 10 percent.

Constant
EC
AFC
ECAFC
gGDP
IND13
IND14
lnTA
gTA
OITA
DEPTA
INVFATA
DRlag
Number of observations
Sargan test
J statistic
p-value
First-differenced residual
correlation
First-order
Second-order
Wald test: c2

Variable

(A) 19832007

Table 6. Dynamic panel-data GMM results for debt ratios

154 Emerging Markets Finance & Trade

SeptemberOctober 2012 Supplement 155

our results showed that debt ratios are countercyclical over the business cycles of the
period 19992007, that is, after the Asian financial crisis, which is in accord with the
findings and conclusions of previous studies on the effect of macroeconomic conditions.
Further, the results revealed that debt ratios increased after the 199798 Asian financial
crisis. In addition, we found that the interaction between macroeconomic conditions and
the Asian financial crisis has a significant, negative effect on debt ratios in these three
industries. Future research may provide new insights into how financial crisis affects
capital structure decisions.
Moreover, we found that debt ratios vary across industries during periods before and
after the 199798 Asian financial crisis. Debt ratios of firms in the electronics industries
are significantly higher than those in the petrochemical and textile industries during the
period before the 199798 Asian financial crisis. This finding is not in line with that of
Bradley et al. (1984). This is probably because of the government industrial policy in
favor of the selective industry during the 1980s and 1990s. However, as a result of the
Taiwanese governments change in the selective industry policy, the debt ratios during the
post-financial crisis period of firms in the electronics industry are significantly higher than
those of the textile industry only. Future research to gather further evidence as regards
the effect of government industrial policy on capital structure is recommended. Finally,
overall, our findings on the effect of firm characteristics on debt ratios in these three
industries are consistent with those of most previous studies on capital structure.
References
Akimov, A., and B. Dollery. 2008. Financial System Reform in Kazakhstan from 1993 to 2006
and Its Socioeconomic Effects. Emerging Markets Finance and Trade 44, no.3: 8197.
Arellano, M., and S. Bond. 1991. Some Tests of Specification for Panel Data: Monte Carlo
Evidence and an Application to Employment Equations. Review of Economic Studies 50,
no.2: 277297.
Baltagi, B.H. 2008. Econometric Analysis of Panel Data. New York: John Wiley & Sons.
Boyd, J., and B. Smith. 1996. The Coevolution of the Real and Financial Sectors in the Growth
Process. World Bank Economic Review 10, no. 2: 371396.
Bradley, M.; G.A. Jarrell; and E.H. Kim. 1984. On the Existence of an Optimal Capital Structure: Theory and Evidence. Journal of Finance 39, no. 3: 857878.
Chang, X., and S. Dasgupta. 2009. Target Behavior and Financing: How Conclusive Is the
Evidence? Journal of Finance 64, no. 4: 17671796.
Chen, J.J. 2004. Determinants of Capital Structure of Chinese-Listed Companies. Journal of
Business Research 57, no. 12: 13411351.
Chu, P.Y.; S. Wu; and S.F. Chiou. 1992. The Determinants of Corporate Capital Structure
Choice: Taiwan Evidence. Journal of Management Science 9, no. 2: 159177 (in Chinese).
Chu, W. 1994. Import Substitution and Export-Led Growth: A Study of Taiwans Petrochemical Industry. World Development 22, no. 5: 781.
. 2003. Taiwans Economy Under Globalization. Taipei: Ton-San (in Chinese).
Council for Economic Planning and Development (CEPD). 2009. Taiwan Business Indicators.
Taipei.
C r+ nigoj, M., and D. Mramor. 2009. Determinants of Capital Structure in Emerging European
Economies: Evidence from Slovenian Firms. Emerging Markets Finance & Trade 45, no.1
(JanuaryFebruary): 7289.
Downs, T.W. 1993. Corporate Leverage and Nondebt Tax Shields: Evidence on CrowdingOut. Financial Review 28, no. 4: 549583.
Feidakis, A., and A. Rovolis. 2007. Capital Structure Choice in European Union: Evidence
from the Construction Industry. Applied Financial Economics 17, no. 12: 9891002.
Flannery, M.J., and K.P. Rangan. 2006. Partial Adjustment Toward Target Capital Structures.
Journal of Financial Economics 79, no.3: 469506.

156 Emerging Markets Finance & Trade

Glen, J., and A. Singh. 2004. Comparing Capital Structures and Rates of Return in Developed
and Emerging Markets. Emerging Markets Review 5, no. 2: 161192.
Hackbarth, D.; J. Miao; and E. Morellec. 2006. Capital Structure, Credit Risk, and Macroeconomic Conditions. Journal of Financial Economics 82, no. 3: 519550.
Harris, M., and A. Raviv. 1991. The Theory of Capital Structure. Journal of Finance 46, no.1:
297355.
Hazak, A. 2009. Companies Financial Decisions Under the Distributed Profit Taxation Regime
of Estonia. Emerging Markets Finance and Trade 45, no. 4: 412.
Hovakimian, A., and G. Li. 2011. In Search of Conclusive Evidence: How to Test for Adjustment to Target Capital Structure. Journal of Corporate Finance 17, no. 1: 3344.
Jensen, M.C. 1986. Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers.
American Economic Review 76, no. 2: 323329.
Kim, W.S., and E.H. Sorensen. 1986. Evidence on the Impact of the Agency Costs of Debt on
Corporate Debt Policy. Journal of Financial and Quantitative Analysis 21, no. 2: 131144.
Korajczyk, R.A., and A. Levy. 2003. Capital Structure Choice: Macroeconomic Conditions and
Financial Constraints. Journal of Financial Economics 68, no. 1: 75109.
Kuo, H., and L. Wang. 2005. The Effect of the Degree of Internationalization on Capital
Structure for Listed Multinational Corporations in Taiwan During the Asian Financial Crisis.
Review of Pacific Basin Financial Markets and Policies 8, no. 3: 447466.
Levy, A., and C. Hennessy. 2007. Why Does Capital Structure Choice Vary with Macroeconomic Conditions? Journal of Monetary Economics 54, no. 6: 15451564.
Miller, M.H. 1977. Debt and Taxes. Journal of Finance 32, no. 2: 261275.
Modigliani, F., and M.H. Miller. 1958. The Cost of Capital, Corporation Finance and the
Theory of Investment. American Economic Review 48, no. 3: 261297.
Nieh, C.; H. Yau; and W. Liu. 2008. Investigation of Target Capital Structure for Electronic
Listed Firms in Taiwan. Emerging Markets Finance and Trade 44, no. 4: 7587.
Ozkan, A. 2001. Determinants of Capital Structure and Adjustment to Long Run Target:
Evidence from UK Company Panel Data. Journal of Business Finance & Accounting 28,
nos.12: 175198.
Qian, Y.; Y. Tian; and T.S. Wirjanto. 2009. Do Chinese Publicly Listed Companies Adjust Their
Capital Structure Toward a Target Level? China Economic Review 20, no. 4: 662676.
Stata. 2009. Longitudinal-Data/Panel-Data Reference Manual Release 11. College Station, TX:
Stata Press.
Titman, S., and R. Wessels. 1988. The Determinants of Capital Structure Choice. Journal of
Finance 43, no. 1: 119.
Wald, J.K. 1999. How Firm Characteristics Affect Capital Structure: An International Comparison. Journal of Financial Research 22, no. 2: 161187.
Windmeijer, F. 2005. A Finite Sample Correction for the Variance of Linear Efficient Two-Step
GMM Estimators. Journal of Econometrics 126, no. 1: 2551.
Wiwattanakantang, Y. 1999. An Empirical Study on the Determinants of the Capital Structure
of Thai Firms. Pacific-Basin Finance Journal 7, nos. 34: 371403.

To order reprints, call 1-800-352-2210; outside the United States, call 717-632-3535.

Copyright of Emerging Markets Finance & Trade is the property of M.E. Sharpe Inc. and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.