
 Start

1834
 Prefix

Corresponding author: Oleksandr Talavera,
tel. (+49) (0)30 89789 407, fax. (+49) (0)30 89789 104, email: otalavera@diw.de, mailing address: DIW Berlin, 10108
Berlin, Germany.
1
1 Introduction
The determinants of capital structure have always attracted considerable attention in
the literature. In their seminal work,
 Exact

Modigliani and Miller (1958)
 Suffix

derive the theoretical result that under the assumption of perfect capitalmarkets, financial and real
decisions are separable so that the firm’s leverage has no effect on the market value
of the firm, nor on its capital investment plans.
 (check this in PDF content)

 Start

2592
 Prefix

show a positive relationship between liquid asset holdings and firms’ investment decisions.1Other studies
show that firm leverage depends on firmspecific characteristics such as cash holdings,
total assets, and the investmenttocapital ratio.2Furthermore, empirical evidence on
the interaction of macroeconomic uncertainty and capital structure indicators is rather
scarce. As an exception,
 Exact

Baum, Caglayan, Ozkan and Talavera(2006)
 Suffix

find a negative
relationship between macroeconomic uncertainty and the crosssectional dispersion of
cashtoasset ratios for US nonfinancial firms. Hence, their study supports the view
that macroeconomic uncertainty is an important factor in firms’ decisionmaking.
 (check this in PDF content)

 Start

3523
 Prefix

The model is based upon anempirically testable
hypothesis regarding the association between the optimal level of debt and uncertainty
arising from macroeconomic or idiosyncratic sources. The model predicts that an increase in either type of uncertainty leads to a decrease in leverage.
1See for example
 Exact

Gilchrist and Himmelberg (1998); Fazzari, Hubbard and Petersen (1988). 2See Shuetrim, Lowe and Morling (1993); Auerbach (1985); Weill (2001)
 Suffix

.
2
To test the model’s predictions, we apply the System GMM estimator (Blundell and
Bond, 1998) to a panel of US nonfinancial firms obtained from the quarterlyCOMPUSTATdatabase over the 1993–2002 period.
 (check this in PDF content)

 Start

3731
 Prefix

The model predicts that an increase in either type of uncertainty leads to a decrease in leverage.
1See for example Gilchrist and Himmelberg (1998); Fazzari, Hubbard and Petersen (1988). 2See Shuetrim, Lowe and Morling (1993); Auerbach (1985); Weill (2001).
2
To test the model’s predictions, we apply the System GMM estimator
 Exact

(Blundell and Bond, 1998) to
 Suffix

a panel of US nonfinancial firms obtained from the quarterlyCOMPUSTATdatabase over the 1993–2002 period. After some screening procedures it includes
more than 31,000 manufacturing firmquarter observations,with about 950 firms per
quarter.
 (check this in PDF content)

 Start

5824
 Prefix

Finally, Section 4 concludes and gives suggestions for further research.
3
2 TheQModel of Firm Value Optimization
2.1 Model Setup
The theoretical model proposed in this paper is based on the firm value optimization
problem and represents a generalization of the standardQmodels of investment by
 Exact

Whited (1992) and Hubbard and Kashyap (1992).
 Suffix

The present value of the firm is
equated to the expected discounted stream ofDt, dividends paid to shareholders, where
βis the discount factor.
Vt(Kt) = max
{It+s,Bt+s}∞s=0
Dt+Et
"∞
X
s=1
βsDt+s
#
,(1)
Kt+1= (1−δ)Kt+It,(2)
Dt= Π(Kt, φt)−C(It, Kt)−It+Bt−Bt−1R(τt)η(Bt−1, Kt, ξt),(3)
Dt≥0,(4)
lim
T→∞
T−1Y
j=t
βj
BT= 0,∀t(5)
The firm maximizes equation (1) subject to three constraints.
 (check this in PDF content)

 Start

7356
 Prefix

We incorporate financial frictions
assuming that riskneutral shareholders require an external premium,η(Bt−1, Kt, ξt),
which depends on firmspecific characteristics such as debt and capital stock as well
as a stochastic shockξt. Similar to
 Exact

Gilchrist and Himmelberg (1998),
 Suffix

we also assume
∂η/∂Bt−1>0: i.e., highly indebted firms must pay an additional premiumto compensate
debtholders for additional costs because of monitoring orhazard problems. Moreover,
∂η/∂Kt<0: i.e., large firms enjoy a lower risk premium.
 (check this in PDF content)

 Start

8641
 Prefix

ExpressionβΘtmay serve as a stochastic timevarying discount
3This is in line with the stylized fact—particularly for large US publiclytraded firms–that equity
finance is a rarelyconsidered option. “The net equity issuances by the US nonfinancial corporate sector
have been negative since the 1980s.”
 Exact

Frank and Goyal (2005),
 Suffix

p. 32. To a first approximation, ignoring
equity finance would seem to be largely in line with observed practice.
5
factor which is equal toβin the absence of financial constraints (λt+1=λt).4
From the firstorder conditions for debt we derive:
Et
h
βΘtRt+1
ηt+1+ηBt+1Bt
<i
= 1.(7)
In the steady stateβEt{(R(τt+1))Θt}=βE{R(τt+1)}= 1,which implies thatηBt+1Bt=
1−ηt+1.
 (check this in PDF content)

 Start

9090
 Prefix

equity finance would seem to be largely in line with observed practice.
5
factor which is equal toβin the absence of financial constraints (λt+1=λt).4
From the firstorder conditions for debt we derive:
Et
h
βΘtRt+1
ηt+1+ηBt+1Bt
<i
= 1.(7)
In the steady stateβEt{(R(τt+1))Θt}=βE{R(τt+1)}= 1,which implies thatηBt+1Bt=
1−ηt+1.Since we assumeηBt+1>0, Btis guaranteed to be positive only ifηt+1≤1.
 Exact

Gilchrist and Himmelberg (1998)
 Suffix

suggest that the risk premium may be negative ifηt+1
is considered as net of tax advantages or agency benefits.
Combining the first order conditions we derive the optimal level for borrowing
Bt=
Et{ΠK,t+1Θt}+ (1−δ)Et{ΘtCI,t+1}−Et{Θtηt+1R(τt+1)}−1/βCI,t
ηBEt{ΘtR(τt+1)}+ηKE{R(τt+1)}
(8)
which allows us to derive
∂Bt
∂τt+1
=
∂Bt
∂Et{R(τt+1)}
∂Et{R(τt+1)}
∂τt+1
<0(9)
Similarly, we construct the relati
 (check this in PDF content)

 Start

9733
 Prefix

{Θtηt+1R(τt+1)}−1/βCI,t
ηBEt{ΘtR(τt+1)}+ηKE{R(τt+1)}
(8)
which allows us to derive
∂Bt
∂τt+1
=
∂Bt
∂Et{R(τt+1)}
∂Et{R(τt+1)}
∂τt+1
<0(9)
Similarly, we construct the relationship between idiosyncratic uncertainty and firm lending. Equation (8) provides a positive relationship betweenexpected profitability and
leverage. The negative relationship between profitabilityand uncertainty is justified in
 Exact

Batra and Ullah (1994)
 Suffix

:
∂Bt
∂φt+1
=
∂Bt
∂Et{Π(φt+1)}
∂Et{Π(φt+1)}
∂φt+1
<0(10)
4For simplicity, we ignore the derivative of the investment adjustment cost function with respect to
the capital stock,CK,t. In our data the mean ofIt/Kt=0.04, and the squared term will be 0.0016 given
thatCK,t= (It/Kt)2.
 (check this in PDF content)

 Start

11117
 Prefix

In order to implement Euler equation estimation we linearize the product ofβt,
ΘtandAt,whereAt= ΠK,t+1+ (1−δ) (CI,t+1+ 1)−R(τt+1)ηK,t+1Bt+1.We utilize
a first–order Taylor approximation around the means. Ignoring constant terms, the
approximation is equal to:5
βtΘtAt=βγΘt+βAt+γβt
whereβis the average discount factor andγdenotes the unconditional mean ofAt.
As in
 Exact

Chirinko (1987) and Hayashi (1982),
 Suffix

we utilize a traditional adjustment cost
function given byC(It, Kt) =α2(It/Kt−νi)2Kt.6The parameterνimight be interpreted as a firmspecific optimal level of investment. The marginal adjustment cost of
5See also Love (2003).
6We acknowledge that there is support in the literature for potentially nonlinear adjustment cost
functions (e.g.
 (check this in PDF content)

 Start

11364
 Prefix

As in Chirinko (1987) and Hayashi (1982), we utilize a traditional adjustment cost
function given byC(It, Kt) =α2(It/Kt−νi)2Kt.6The parameterνimight be interpreted as a firmspecific optimal level of investment. The marginal adjustment cost of
5See also
 Exact

Love (2003).
 Suffix

6We acknowledge that there is support in the literature for potentially nonlinear adjustment cost
functions (e.g., Abel and Eberly (2002). As this feature is not the focus of our study, we make the
standard assumption of quadratic adjustment costs.
7
investment of a firmiat timetis given by:
CI,it=α
>
Iit
T Ait
−νi
(11)
whereT Ait, a proxy for capitalKit, measures total assets of firmiat timet.
 (check this in PDF content)

 Start

11491
 Prefix

As in Chirinko (1987) and Hayashi (1982), we utilize a traditional adjustment cost
function given byC(It, Kt) =α2(It/Kt−νi)2Kt.6The parameterνimight be interpreted as a firmspecific optimal level of investment. The marginal adjustment cost of
5See also Love (2003). 6We acknowledge that there is support in the literature for potentially nonlinear adjustment cost
functions (e.g.,
 Exact

Abel and Eberly (2002).
 Suffix

As this feature is not the focus of our study, we make the
standard assumption of quadratic adjustment costs.
7
investment of a firmiat timetis given by:
CI,it=α
>
Iit
T Ait
−νi
(11)
whereT Ait, a proxy for capitalKit, measures total assets of firmiat timet.
 (check this in PDF content)

 Start

12642
 Prefix

The level of financing constraint for a representative firmiat timet, Θit, is a function
of their stock of cash and level of debt:
Θit=a0i+a1Cashit+a2Bi,t−1(14)
whereCashitis the cash and cash equivalent,Bitis the level of debt anda0iis a
firmspecific measure of financial constraints. Debt generates interest and principal
7The discussion in
 Exact

Gilchrist and Himmelberg (1998)
 Suffix

suggeststhat a salesbased measure of the
marginal profit of capital is more desirable comparing to operating income measure.
8
obligations and increases the probability of financial distress, while the availability of
liquid assets relaxes the external finance constraint (see also Hubbard, Kashyap and
Whited (1995); Almeida, Campello and Weisbach (2004); Gilchrist and Himmelberg
(1998)).
 (check this in PDF content)

 Start

12953
 Prefix

Debt generates interest and principal
7The discussion in Gilchrist and Himmelberg (1998) suggeststhat a salesbased measure of the
marginal profit of capital is more desirable comparing to operating income measure.
8
obligations and increases the probability of financial distress, while the availability of
liquid assets relaxes the external finance constraint (see also
 Exact

Hubbard, Kashyap and Whited (1995); Almeida, Campello and Weisbach (2004); Gilchrist and Himmelberg (1998)).
 Suffix

Finally, the base interest rateR(τt) is assumed to be a linear function of macroeconomic uncertainty and the index of leading indicators, a proxy of the state of the
macroeconomy:
R(τt) =ω1τt+ω2Leadingt(15)
The resulting empirical specification is:8
Bit
T Ait
=β0+β1
Bit−1
T Ait−1
+β2
Cashit
T Ait
+β3
Sit
T Ait
+β4
Iit+1
T Ait+1
+β5
Iit
T Ait
(16)
+β6τt−1+β7φi,t−1+β8Leadinct−1+fi+Indi+eit
AsCOMPUST
 (check this in PDF content)

 Start

14771
 Prefix

The main aim of our study is to investigate whether robust results are obtained for
these hypotheses relating to uncertainty measures and not to identify the coefficients of
the structural model.9
2.3 Identifying Uncertainty
The macroeconomic uncertainty identification approach resembles that of
 Exact

Baum et al. (2006).
 Suffix

Firms’ debt decisions depend on anticipation of future profits and investment.
The difficulty of evaluating the optimal amount of debt issuing increases with the level
of macroeconomic uncertainty.
The literature suggests various methods to obtain a proxy for macroeconomic uncertainty.
 (check this in PDF content)

 Start

15103
 Prefix

The difficulty of evaluating the optimal amount of debt issuing increases with the level
of macroeconomic uncertainty.
The literature suggests various methods to obtain a proxy for macroeconomic uncertainty. In our investigation, as in
 Exact

Driver, Temple and Urga (2005) and Byrne and Davis (2005),
 Suffix

we use aGARCHmodel to proxy for macroeconomic uncertainty. We
believe that this approach is more appropriate compared to alternatives such as proxies
obtained from moving standard deviations of the macroeconomic series (e.g.
 (check this in PDF content)

 Start

15386
 Prefix

In our investigation, as in Driver, Temple and Urga (2005) and Byrne and Davis (2005), we use aGARCHmodel to proxy for macroeconomic uncertainty. We
believe that this approach is more appropriate compared to alternatives such as proxies
obtained from moving standard deviations of the macroeconomic series (e.g.,
 Exact

Ghosal and Loungani (2000))
 Suffix

or surveybased measures based on the dispersion of forecasts (e.g.,
Graham and Harvey (2001), Schmukler, Mehrez and Kaufmann (1999)). While the
9It is possible to show that allβs are functions of model parameters, but it is not possible toidentify
every parameter of the structural model with reasonable assumptions.
10
former approach suffers from substantial serial correlation problems in the co
 (check this in PDF content)

 Start

15482
 Prefix

We
believe that this approach is more appropriate compared to alternatives such as proxies
obtained from moving standard deviations of the macroeconomic series (e.g., Ghosal and Loungani (2000)) or surveybased measures based on the dispersion of forecasts (e.g.,
 Exact

Graham and Harvey (2001), Schmukler, Mehrez and Kaufmann (1999)).
 Suffix

While the
9It is possible to show that allβs are functions of model parameters, but it is not possible toidentify
every parameter of the structural model with reasonable assumptions.
10
former approach suffers from substantial serial correlation problems in the constructed
series the latter potentially contains sizable measurement errors.
 (check this in PDF content)

 Start

16507
 Prefix

The specifics of
theGARCHmodel are provided in Table 1. The estimated conditional variance series is
then employed in a revised version of equation (16).
One can employ different proxies to capture firmspecific risk. For instance,
 Exact

Bo and Lensink (2005)
 Suffix

use three measures: stock price volatility,estimated as the difference
between the highest and the lowest stock price normalized bythe lowest price; volatility
of sales measured by the coefficient of variation of sales overa seven–year window; and
the volatility of number of employees estimated similarly to volatility of sales.
 (check this in PDF content)

 Start

16855
 Prefix

For instance, Bo and Lensink (2005) use three measures: stock price volatility,estimated as the difference
between the highest and the lowest stock price normalized bythe lowest price; volatility
of sales measured by the coefficient of variation of sales overa seven–year window; and
the volatility of number of employees estimated similarly to volatility of sales.
 Exact

Bo (2002)
 Suffix

employs a slightly different approach, setting up the forecasting AR(1) equation for the
underlying uncertainty variable driven by sales and interest rates. The unpredictable
part of the fluctuations, the estimated residuals, are obtained from that equation and
their threeyear moving average standard deviation is computed.
 (check this in PDF content)

 Start

21736
 Prefix

, paper products, printing and publishing,
chemicals, petroleum and coal products, rubber and plastics, and leather products makers.
13
3.2 Empirical results
We estimate Equation (17) using the lagged conditional variance of the index of leading
indicators as the proxy for macroeconomic uncertainty. Results for all manufacturing
firms are given in Table 4. These specifications represent the
 Exact

Blundell and Bond (1998)
 Suffix

onestep System GMM estimator employing the first differences transformation.12As
instruments we useB/T At−3toB/T At−5,CASH/T At−2toCASH/T At−5,I/T At−2
toI/T At−5, andS/T At−2toS/T At−5for the equations in differences and ∆S/T At−1,
∆CASH/T At−1, and ∆I/T At−1for the equations in levels.
 (check this in PDF content)

 Start

26719
 Prefix

Durable goods makers, highliquidity firms, highly leveraged firms and small firms also
exhibit sensitivity to firmspecific volatility.
From the policy perspective, we suggest that macroeconomicuncertainty has an effect
on nonfinancial firms’ capital structure which in turn affectstheir dynamics of investment. Other studies (see
 Exact

Bernanke and Gertler (1989))
 Suffix

have shown that balance sheet
shocks may affect the amplitude of investment cycles in a simple neoclassical model.
Moreover, in many countries monetary policy tends to be persistent in the direction
of change of the monetary instrument, with rare reversals (perhaps reflecting central
banks’ interest rate smoothing objectives).
 (check this in PDF content)