Professional Documents
Culture Documents
175
176
OZKAN
177
The findings of this paper suggest that firms have target leverage
ratios and they adjust to the target ratio relatively fast. This
implies that the costs of being away from their target ratios and
the costs of adjustment are both important for firms. Another
major conclusion is that current liquidity and profitability of
firms exert a negative impact on their borrowing decisions
whereas there is a positive relation between past profitability and
debt ratio. Other firm-specific variables which appear to
influence leverage decision are non-debt tax shields and growth
opportunities of firms. There is only limited evidence that firm
size exerts an impact on capital structure decisions.
The rest of the paper is organised as follows. In the next
section we give a summary of the findings of the previous
empirical studies. Section 3 discusses the potential determinants
of target leverage ratios used in the study. In Section 4 the
empirical model is presented and a brief description of our data
is given. In Section 5 we present the estimation results and
discuss the robustness of results. The final section provides
concluding remarks.
178
OZKAN
179
In what follows we discuss and motivate the choice of firmspecific characteristics which are thought to influence the
borrowing decisions of firms.
(i) Size
It has been suggested by a number of authors that firm size is
positively related to leverage ratio. The rationale for this belief is
the evidence provided by Warner (1977) and Ang et al. (1982)
that the ratio of direct bankruptcy costs to the firm value
decreases as the value increases, suggesting that the impact of
Blackwell Publishers Ltd 2001
180
OZKAN
181
182
OZKAN
(v) Liquidity
Liquidity ratios may have a mixed impact on the capital structure
decision. First, firms with higher liquidity ratios might support a
relatively higher debt ratio due to greater ability to meet shortterm obligations when they fall due. This would imply a positive
relationship between a firm's liquidity position and its debt ratio.
On the other hand, firms with greater liquid assets may use these
assets to finance their investments. Therefore the firm's liquidity
position should exert a negative impact on its leverage ratio.
Moreover, as Prowse (1990) argues, the liquidity of the
company's assets can be used to show the extent to which these
assets can be manipulated by shareholders at the expense of
bondholders. We use the ratio of current assets to current
liabilities as a proxy for the liquidity of the firm's assets.
The data source for our analysis is the Datastream database which
provides both accounting data on firms and the market value of
equity. The sample data was constructed in the following manner.
First, firms which operate in the financial sector (banks, insurance
companies and investment trusts) and utilities (companies
providing a public service such as electricity, gas and telephone)
were excluded. All firms with any missing observations for any
variable in the model during the sample period were dropped.
Firms were then chosen that had at least five continuous time
series observations during the period 19841996, resulting in a
final sample of 390 firms and 4,132 observations.
In the next section we report alternative estimates of the
following specification
Dit Di;t1 ; Sizeit ; Liquidityit ; Ndtsit ; Profitit ; Growthit
i t "it
183
184
OZKAN
5. RESULTS
Columns (1) and (2) of Table 1 present GMM and AndersonHsiao (AH) type estimates respectively, both in first differenced
form. Column (3) gives OLS estimates of leverage equation (1)
in levels. The estimation period is 198696 for both the GMM
and the AH estimates as a result of losing two cross-sections in
constructing one lag for each variable and taking first
differences. In both models, the fixed effects are eliminated by
first differencing and all the variables except the lagged
dependent variable are treated as exogenous. However, in model
(3) which gives the OLS estimates, the lagged dependent variable
is also treated as exogenous and the unobservable firm-specific
fixed effects remain. In all models the sample contains 390 firms
and 4,132 observations although usable observations vary
according to the estimation method. We report six test statistics:
(1) First order autocorrelation of residuals which is
asymptotically distributed as standard normal N(0,1) under the
null of no serial correlation; (2) Second order autocorrelation of
residuals which is distributed as standard normal N(0,1) under
the null of no serial correlation; (3) Wald test 1 is a Wald test of
joint significance of the estimated coefficients which is
asymptotically distributed as chi-square under the null of no
relationship; (4) Wald test 2 is a Wald test of the joint significance
of the time dummies; (5) Wald test 3 is a Wald test of the joint
significance of the industry dummies; and (6) Sargan test of
overidentifying restrictions which is asymptotically distributed as
chi-square under the null of instrument validity. All estimations
were carried out using the DPD program written in GAUSS
(Arellano and Bond, 1988).
Turning to the estimation results and comparing the GMM and
AH estimates it can be seen that the coefficient estimates
(including the lagged dependent variable) under AH are poorly
determined. The AH results reveal substantially larger variance
than that associated with the GMM, suggesting a loss in efficiency
compared to the GMM estimates. This is consistent with the
findings of Arellano and Bond (1991). There is also evidence that
the OLS level specification is inappropriate to estimate our
dynamic model. First, the serial correlation tests reveal that the
assumption of serially uncorrelated errors is violated and this
Blackwell Publishers Ltd 2001
185
Table 1
Alternative Estimates of Target Capital Structure
Dependent Variable: Dit
Independent
Variables
Di;t1
Sizeit
Sizeit1
Liquidityit
Liquidityit1
Ndtsit
Ndtsit1
Profitit
Profitit1
Growthit
Growthit1
Correlation 1
Correlation 2
Wald test 1 (df)
Wald test 2 (df)
Wald test 3 (df)
Sargan test (df)
(1)
GMM
(2)
AH
(3)
OLS
0.590***
(0.028)
0.077***
(0.009)
0.032***
(0.008)
0.055***
(0.011)
0.026***
(0.005)
0.059
(0.111)
0.020
(0.159)
0.244***
(0.026)
0.061**
(0.026)
0.024***
(0.005)
0.008**
(0.004)
0.295**
(0.132)
0.075***
(0.012)
0.012
(0.012)
0.035**
(0.015)
0.009*
(0.006)
0.064
(0.157)
0.190
(0.205)
0.225***
(0.037)
0.005
(0.044)
0.020***
(0.007)
0.006
(0.006)
0.781***
(0.015)
0.050***
(0.010)
0.047***
(0.010)
0.035***
(0.009)
0.030***
(0.008)
0.010
(0.124)
0.112
(0.133)
0.238***
(0.032)
0.133***
(0.027)
0.009*
(0.005)
0.003
(0.005)
11.560
1.258
979.68 (11)
124.40 (11)
88.06 (65)
3.966
0.547
2101.89 (11)
49.91 (11)
5.241
1.503
3772.02 (11)
46.74 (10)
5.884 (4)
Notes:
The dependent variable d is defined as the ratio of total debt to total assets. `Size' is the
natural logarithm of sales. `Liquidity' is the ratio of current assets to current liabilities.
`Ndts' is the ratio of annual depreciation expense to total assets. `Profits' is the ratio of
EBITD to total assets. `Growth' is the ratio of book value of total assets minus the book
value of equity plus the market value of equity to book value of total assets.
Time dummies are included in all models. Industry dummies are included only in model
3. Asymptotic standard errors robust to heteroscedasticity are reported in parentheses.
Model 1 gives the GMM estimates in first differences where Di;t2 , and further lags are
used as instruments. In Model 2 Anderson-Hsiao type estimates in first differences are
reported where Di;t2 is used as instrument for Di;t1 . Model 3 gives OLS estimates in
levels. ***, ** and * indicate coefficient is significant at the 1%, 5% and 10% levels,
respectively.
186
OZKAN
187
Table 2
GMM Estimates of Target Capital Structure
Dependent Variable: Dit
Independent
Variables
Di;t1
Sizeit
Sizeit1
Liquidityit
Liquidityit1
Ndtsit
Ndtsit1
Profitit
Profitit1
Growthit
Growthit1
Correlation 1
Correlation 2
Wald testr 1 (df)
Wald test 2 (df)
Sargan test (df)
(1)
GMM
(2)
GMM
(3)
GMM
0.431***
(0.036)
0.034
(0.024)
0.053**
(0.024)
0.025***
(0.006)
0.002
(0.003)
1.208***
(0.414)
0.240
(0.300)
0.457***
(0.096)
0.123***
(0.041)
0.044***
(0.014)
0.047***
(0.012)
0.485***
(0.035)
0.029
(0.024)
0.049**
(0.025)
0.019***
(0.006)
0.002
(0.004)
1.151***
(0.444)
0.090
(0.296)
0.375***
(0.092)
0.104**
(0.041)
0.040***
(0.014)
0.043***
(0.012)
0.443***
(0.034)
0.024*
(0.013)
0.020***
(0.005)
1.071***
(0.367)
0.495***
(0.093)
0.121***
(0.042)
0.040***
(0.013)
0.047***
(0.012)
11.478
0.192
477.73 (11)
55.314 (11)
63.25 (54)
11.492
0.996
502.84 (11)
55.07 (11)
70.36 (54)
11.666
0.758
468.03 (8)
66.08 (11)
66.35 (58)
Notes:
The dependent variable D is defined as the ratio of total debt to total assets in Models (1)
and (3) and the ratio of total debt plus preferred capital to total assets in Model (2). `Size'
is the natural logarithm of sales. `Liquidity' is the ratio of current assets to current
liabilities. `Ndts' is the ratio of annual depreciation expense to total assets. `Profit' is the
ratio of EBITD to total assets. `Growth' is the ratio of book value of total assets minus the
book value of equity plus the market value of equity to book value of total assets.
In all models Di;t2 , Liquidityi;t2 , Ndtsi;t2 , Profiti;t2 , Growthi;t2 are used as instruments.
The estimation period is 198696 for all models.
See also notes to Table 1.
188
OZKAN
189
190
OZKAN
191
192
OZKAN
6. CONCLUSIONS
193
APPENDIX
Empirical Specification
We investigate the role of adjustment costs by adopting a partial
adjustment model. Suppose that the target long-term debt ratio
of firms is taken to be a function of several variables:
X
Dit
k xkit "i t
A1
k
194
OZKAN
A2
where 0 < < 1, Dit is the actual debt ratio and Dit is the target
debt ratio of firm i at time t. Dit Di;t1 can be interpreted as
the target change whereas only a fraction of the target change
is achieved, which is equal to Dit Di;t1 . Combining (A1) and
(A2) yields:
X
Dit 1 Di;t1
k xkit it
A3
k
k xkit uit
A4
k1
Number of Companies
24
17
32
31
34
33
15
15
189
195
Table A2
The Number of Companies in Each Year
Year
Number of Companies
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
223
237
253
285
316
337
362
370
370
359
350
341
329
Table A3
Descriptive Statistics
Variable
Leverage
Leveraget1
Sizet1
Growth
Growtht1
Ndts
Profit
Profitt1
Liquidity
0.168
0.167
11.234
0.288
0.274
0.036
0.103
0.107
1.641
(0.106)
(0.105)
(1.807)
(0.504)
(0.500)
(0.020)
(0.087)
(0.079)
(0.756)
First Differences
0.001
0.002
0.116
0.013
0.010
0.001
0.004
0.003
0.004
(0.071)
(0.071)
(0.247)
(0.302)
(0.312)
(0.012)
(0.078)
(0.070)
(0.467)
NOTES
1
Bennett and Donnelly (1993) and Walsh and Ryan (1997) use the ratio of
unprovided deferred tax to total assets as a proxy for non-debt tax shields.
Unfortunately, in our data set there were too many missing observations for
these variables so that we could not use it as an alternative proxy for nondebt tax shields.
The GMM estimator allows the instruments used in each period to increase
as one moves through the panel, whereas the AH type estimator uses only
Di;t2 to instrument Di;t1 . The set of valid instruments change
depending upon the assumption concerning the correlation between xikt
196
OZKAN
and uit . It is suggested that the valid instruments for period t for the
equation in first differences will be zit Di1 ; . . . ; Di;t2 ; xik1 ; . . . ; xik;t1
under the assumption that uit is serially uncorrelated, and xikt is predetermined. That is, Exkit uis 6 0 for s < t and zero otherwise. If, on the
other hand, xikt is strictly exogenous, i.e. Exikt uis 0 for all t,s, then all x's
are valid instruments. In this case zit becomes zit Dit ; . . . ; Dis ; xik1 ; . . . ; xikT
where s 1; . . . ; T 2.
One should be cautious in making comparison between alternative
estimates at this stage because this may be obscured by the likely
endogeneity of firm-specific characteristics. To explore this possibility, we
treated all variables as endogenous and derive the results accordingly. This
did not change the basic results in that the coefficient of the lagged
dependent variable is significantly biased upwards in OLS specification and
the coefficient estimates of the AH specification are poorly determined
compared to the GMM estimates.
Note that, if uit are serially uncorrelated in the model in levels of first
differences, transformation induces first-order serial correlation in the
differenced residuals, but second order serial correlation should be absent.
That is, as explained in Arellano and Bond (1991), Euit uit1 need not be
zero, but the consistency of the GMM estimators relies heavily on the
assumption that Euit uit2 0.
REFERENCES
Anderson, T.W. and C. Hsiao (1982), `Formulation and Estimation of Dynamic
Models Using Panel Data', Journal of Econometrics, Vol. 18, pp. 4782.
Ang, J.S., J.H.Chua, and J.J. McConnell (1982), `The Administrative Costs of
Corporate Bankruptcy: A Note', Journal of Finance, Vol. 37, pp. 33748.
Arellano, M. and S. Bond (1988), `Dynamic Panel Data Estimation Using DPD
A Guide for Users', Institute for Fiscal Studies, Working Paper 88/15
(London).
________ ________ (1991), `Some Tests of Specification for Panel Data: Monte
Carlo Evidence and an Application to Employment Equations', The Review
of Economics Studies, Vol. 58, pp. 27797.
Auerbach, A.J. (1985), `Real Determinants of Corporate Leverage', in Friedman
(ed.), Corporate Capital Structure in the United States (University of Chicago
Press, Chicago).
Barclay, M.J. and C.W. Smith (1995), `The Maturity Structure of Corporate
Debt', Journal of Finance, Vol. 50, pp. 60932.
Bennett, M. and R. Donnelly (1993), `The Determinants of Capital Structure:
Some UK Evidence', British Accounting Review, Vol. 25, pp. 4359.
Blundell, R., S. Bond, M. Devereux, and F. Schiantarelli (1992), `Investment
and Tobin's Q: Evidence from Company Panel Data', Journal of Econometrics,
Vol. 51, pp. 23357.
Bowen, R.M., L.A. Daley, and C.C. Huber (1982), `Evidence on the Existence
and Determinants of Inter-Industry Differences of Leverage', Financial
Management, Vol.11, pp. 1020.
Bradley, M., A. Jarrell and E. Kim (1984), `On the Existence of an Optimal
Capital Structure: Theory and Evidence', Journal of Finance, Vol. 39, pp.
85780.
Blackwell Publishers Ltd 2001
197
198
OZKAN
Behaviour in the U.S. and Japan', Journal of Financial Economics, Vol. 27, pp.
4366.
Rajan R.G. and L. Zingales (1995), `What Do We Know About Capital Structure?
Some Evidence from International Data', Journal of Finance, Vol. 50, pp.
142160.
Scott, J.H. (1976), `A Theory of Optimal Capital Structure', Bell Journal of
Economics, Vol. 7, pp. 3354.
Shyam-Sunder, L. and S.C. Myers (1999), `Testing Static Tradeoff Against
Pecking Order Models of Capital Structure', Journal of Financial Economics,
Vol. 51, pp. 21944.
Smith, C.W. and R.L. Watts (1992), `The Investment Opportunity Set and
Corporate Financing, Dividend, and Compensation Policies', Journal of
Financial Economics, Vol. 32, pp. 26392.
Taggart, R.A. (1977), `A Model of Corporate Financing Decisions', Journal of
Finance, Vol. 32, pp. 146784.
Titman, S. and R. Wessels (1988), `The Determinants of Capital Structure
Choice', Journal of Finance, Vol. 43, pp. 119.
Walsh, E.J. and J. Ryan (1997), `Agency and Tax Explanations of Security
Issuance Decisions', Journal of Business Finance & Accounting, Vol. 24, Nos 7
& 8 (September), pp. 94361.
Warner, J.B. (1977), `Bankruptcy Costs: Some Evidence', Journal of Finance, Vol.
32, pp. 33748.
Whited, T. (1992), `Debt, Liquidity Constraints, and Corporate Investment:
Evidence from Panel Data', Journal of Finance, Vol. 47, pp. 142560.
Williamson, O. (1988), `Corporate Finance and Corporate Governance', Journal
of Finance, Vol. 43, pp. 56791.