
 Start

3503
 Prefix

We specifically consider the effects of three different forms of uncertainty
on firms’ cost of external funds, and thus on their investment behavior:Own
(intrinsic) uncertainty, derived from firms’ stock returns;M arket(extrinsic)
1See
 Exact

Hartman (1972), Abel (1983), Bernanke (1983), Craine (1989), Dixit and Pindyck (1994), Caballero (1999).
 Suffix

2
uncertainty, driven by S&P 500 index returns,2and the relations between
intrinsic and extrinsic uncertainty. To capture the latter effect, we introduce
acovarianceterm (ourCAPMbased risk measure) and allow the data to
determine the differential impact of each of these components on the firm’s
capital investment behavior.
 (check this in PDF content)

 Start

4371
 Prefix

We employ annual firmlevel U.S. manufacturing sector data obtained
fromCOMPUSTATand match it to firmlevel daily financial data fromCRSP
over the 1984–2003 period. Daily stock returns and market index returns are
utilized to compute intrinsic and extrinsic uncertainty via a method based on
 Exact

Merton (1980)
 Suffix

from the intraannual variations in stock returns and aggregate financial market series. This approach provides a more representative
measure of the perceived volatility while avoiding potential problems: for
instance, the high persistence of shocks or low correlation in volatility.
 (check this in PDF content)

 Start

4922
 Prefix

In
that respect, our study improves upon much of the literature in its method
of using highfrequency data to quantify volatility evaluated at a lower frequency.3
The results of the paper can be summarized as follows. In contrast to
earlier research such as
 Exact

Leahy and Whited (1996),
 Suffix

we find a significant role
for each uncertainty measure while factors such as cash flow and the debt
ratio maintain their significance in explaining firm investment behavior.4
Our empirical model evaluates how the effects of uncertainty on investment
may be strengthened or weakened by the firm’s current financial condition.
 (check this in PDF content)

 Start

5727
 Prefix

Furthermore, its effects through cash
flow, along with those of intrinsic uncertainty, are significant but vary in
sign over the range of cash flow values. Interestingly, we find that while
2In this paper we use the termsOwn, idiosyncratic and intrinsic uncertainty interchangeably. Likewise,Marketis taken as synonymous with extrinsic uncertainty.
3
 Exact

Leahy and Whited (1996),
 Suffix

Bloom, Bond and Van Reenen (2007), Bond and Cummins
(2004) have also utilized daily stock returns to compute firmlevel uncertainty. However,
the methodology they used to generate a proxy for uncertainty is different from ours.
4An exception are the findings of Baum, Caglayan and Talavera (2008), which also
display a significant role for similar measures of uncertainty.
 (check this in PDF content)

 Start

5787
 Prefix

Interestingly, we find that while
2In this paper we use the termsOwn, idiosyncratic and intrinsic uncertainty interchangeably. Likewise,Marketis taken as synonymous with extrinsic uncertainty.
3Leahy and Whited (1996), Bloom, Bond and Van Reenen (2007),
 Exact

Bond and Cummins (2004)
 Suffix

have also utilized daily stock returns to compute firmlevel uncertainty. However,
the methodology they used to generate a proxy for uncertainty is different from ours.
4An exception are the findings of Baum, Caglayan and Talavera (2008), which also
display a significant role for similar measures of uncertainty.
 (check this in PDF content)

 Start

6011
 Prefix

Likewise,Marketis taken as synonymous with extrinsic uncertainty.
3Leahy and Whited (1996), Bloom, Bond and Van Reenen (2007), Bond and Cummins (2004) have also utilized daily stock returns to compute firmlevel uncertainty. However,
the methodology they used to generate a proxy for uncertainty is different from ours.
4An exception are the findings of
 Exact

Baum, Caglayan and Talavera (2008),
 Suffix

which also
display a significant role for similar measures of uncertainty. However, their analysis does
not consider interactions of uncertainty with cash flow.
3
the effects ofOwnuncertainty through cash flows on firms’ fixed investment
is positive, that ofMarketuncertainty is negative.
 (check this in PDF content)

 Start

6355
 Prefix

However, their analysis does
not consider interactions of uncertainty with cash flow.
3
the effects ofOwnuncertainty through cash flows on firms’ fixed investment
is positive, that ofMarketuncertainty is negative. In contrast to those of
 Exact

Bloom et al. (2007),
 Suffix

our findings suggest that (although the two models
differ) different types of uncertainty can enhance or impair fixed investment
by themselves or through cash flow, potentially clouding the relationship
between investment and uncertainty (Boyle and Guthrie (2003)).
 (check this in PDF content)

 Start

6610
 Prefix

In contrast to those of
Bloom et al. (2007), our findings suggest that (although the two models
differ) different types of uncertainty can enhance or impair fixed investment
by themselves or through cash flow, potentially clouding the relationship
between investment and uncertainty
 Exact

(Boyle and Guthrie (2003)).
 Suffix

We also
show that the impact of cash flow on capital investment changes as the
underlying uncertainty varies.
The rest of the paper is constructed as follows. Section 2, though not
comprehensive given the vast literature on capital investment, provides a
brief survey of the empirical literature discussing the effects of uncertainty
on investment.
 (check this in PDF content)

 Start

7650
 Prefix

Fluctuations in aggregate investment can arise from various sources
of uncertainty. For instance, many researchers have studied the impact of
exchange rate uncertainty on aggregate or industry level investment behavior. To that end
 Exact

Goldberg (1993)
 Suffix

shows that exchange rate uncertainty has
a weak negative effect on investment spending. Campa and Goldberg (1995)
find no significant impact of exchange rate volatility on investment. Darby,
Hallett, Ireland and Piscatelli (1999) provide evidence that exchange rate
uncertainty may or may not depress investment, while Serven (2003) unearths a highly significant negative impact of real exchange rate
 (check this in PDF content)

 Start

7753
 Prefix

For instance, many researchers have studied the impact of
exchange rate uncertainty on aggregate or industry level investment behavior. To that end Goldberg (1993) shows that exchange rate uncertainty has
a weak negative effect on investment spending.
 Exact

Campa and Goldberg (1995)
 Suffix

find no significant impact of exchange rate volatility on investment. Darby,
Hallett, Ireland and Piscatelli (1999) provide evidence that exchange rate
uncertainty may or may not depress investment, while Serven (2003) unearths a highly significant negative impact of real exchange rate uncertainty
on private investment in a sample of developing countries.
 (check this in PDF content)

 Start

7848
 Prefix

To that end Goldberg (1993) shows that exchange rate uncertainty has
a weak negative effect on investment spending. Campa and Goldberg (1995) find no significant impact of exchange rate volatility on investment.
 Exact

Darby, Hallett, Ireland and Piscatelli (1999)
 Suffix

provide evidence that exchange rate
uncertainty may or may not depress investment, while Serven (2003) unearths a highly significant negative impact of real exchange rate uncertainty
on private investment in a sample of developing countries.
 (check this in PDF content)

 Start

7981
 Prefix

Campa and Goldberg (1995) find no significant impact of exchange rate volatility on investment. Darby, Hallett, Ireland and Piscatelli (1999) provide evidence that exchange rate
uncertainty may or may not depress investment, while
 Exact

Serven (2003)
 Suffix

unearths a highly significant negative impact of real exchange rate uncertainty
on private investment in a sample of developing countries.
Many other researchers have investigated the importance of uncertainty
arising from output, prices (inflation), taxes and interest rates.
 (check this in PDF content)

 Start

8270
 Prefix

Piscatelli (1999) provide evidence that exchange rate
uncertainty may or may not depress investment, while Serven (2003) unearths a highly significant negative impact of real exchange rate uncertainty
on private investment in a sample of developing countries.
Many other researchers have investigated the importance of uncertainty
arising from output, prices (inflation), taxes and interest rates.
 Exact

Driver and Moreton (1991)
 Suffix

conclude that while a proxy for uncertainty driven from
4
output growth has a negative long–run effect on aggregate investment, the
measure of uncertainty obtained from inflation has none. Calcagnini and
Saltari (2000) suggest that while demand uncertainty has a significant negative effect on investment, interest rate uncertainty has none.
 (check this in PDF content)

 Start

8481
 Prefix

Driver and Moreton (1991) conclude that while a proxy for uncertainty driven from
4
output growth has a negative long–run effect on aggregate investment, the
measure of uncertainty obtained from inflation has none.
 Exact

Calcagnini and Saltari (2000)
 Suffix

suggest that while demand uncertainty has a significant negative effect on investment, interest rate uncertainty has none. Huizinga (1993)
reports a negative effect on investment for uncertainty proxies obtained from
wages and raw materials prices, but a positive effect for a proxy obtained
from output prices.
 (check this in PDF content)

 Start

8634
 Prefix

Driver and Moreton (1991) conclude that while a proxy for uncertainty driven from
4
output growth has a negative long–run effect on aggregate investment, the
measure of uncertainty obtained from inflation has none. Calcagnini and Saltari (2000) suggest that while demand uncertainty has a significant negative effect on investment, interest rate uncertainty has none.
 Exact

Huizinga (1993)
 Suffix

reports a negative effect on investment for uncertainty proxies obtained from
wages and raw materials prices, but a positive effect for a proxy obtained
from output prices. Ferderer (1993) captures a measure of uncertainty on
long term bonds using the term structure of interest rates and finds a negative impact on aggregate investment.
 (check this in PDF content)

 Start

8820
 Prefix

Calcagnini and Saltari (2000) suggest that while demand uncertainty has a significant negative effect on investment, interest rate uncertainty has none. Huizinga (1993) reports a negative effect on investment for uncertainty proxies obtained from
wages and raw materials prices, but a positive effect for a proxy obtained
from output prices.
 Exact

Ferderer (1993)
 Suffix

captures a measure of uncertainty on
long term bonds using the term structure of interest rates and finds a negative impact on aggregate investment. Hurn and Wright (1994) find that the
linkage between oil price variability and the decision to develop an oil field
(more specifically the North Sea oil field) is not significant.
 (check this in PDF content)

 Start

8985
 Prefix

Huizinga (1993) reports a negative effect on investment for uncertainty proxies obtained from
wages and raw materials prices, but a positive effect for a proxy obtained
from output prices. Ferderer (1993) captures a measure of uncertainty on
long term bonds using the term structure of interest rates and finds a negative impact on aggregate investment.
 Exact

Hurn and Wright (1994)
 Suffix

find that the
linkage between oil price variability and the decision to develop an oil field
(more specifically the North Sea oil field) is not significant. Pindyck and
Solimano (1993) use the variance in the marginal revenue product of capital
as a proxy for uncertainty to study an implication of irreversible investment
models to find the effects of uncertainty on the investment trigger.
 (check this in PDF content)

 Start

9163
 Prefix

Ferderer (1993) captures a measure of uncertainty on
long term bonds using the term structure of interest rates and finds a negative impact on aggregate investment. Hurn and Wright (1994) find that the
linkage between oil price variability and the decision to develop an oil field
(more specifically the North Sea oil field) is not significant.
 Exact

Pindyck and Solimano (1993)
 Suffix

use the variance in the marginal revenue product of capital
as a proxy for uncertainty to study an implication of irreversible investment
models to find the effects of uncertainty on the investment trigger.
 (check this in PDF content)

 Start

9758
 Prefix

Edmiston (2004) investigates the role of tax uncertainty on investment and finds
a significant negative effect between the two.5
Turning now to research which has used firm level data, we also see
several studies employing measures of uncertainty that emerge from movements in exchange rates, output, demand, firmspecific liquidity, inflation or
a CAPM framework.6
 Exact

Brainard, Shoven and Weiss (1980)
 Suffix

find that aCAPMbased risk measure yields mixed results on the linkages between investment
and their uncertainty measure. Ghosal and Loungani (1996) report a negative role of output uncertainty on investment.
 (check this in PDF content)

 Start

9913
 Prefix

two.5
Turning now to research which has used firm level data, we also see
several studies employing measures of uncertainty that emerge from movements in exchange rates, output, demand, firmspecific liquidity, inflation or
a CAPM framework.6Brainard, Shoven and Weiss (1980) find that aCAPMbased risk measure yields mixed results on the linkages between investment
and their uncertainty measure.
 Exact

Ghosal and Loungani (1996)
 Suffix

report a negative role of output uncertainty on investment. Leahy and Whited (1996),
using risk measures constructed from stock return data, argue that uncertainty exerts a strong negative effect on investment and point out that uncertainty affects investment directly rather than through covariances.
 (check this in PDF content)

 Start

10001
 Prefix

employing measures of uncertainty that emerge from movements in exchange rates, output, demand, firmspecific liquidity, inflation or
a CAPM framework.6Brainard, Shoven and Weiss (1980) find that aCAPMbased risk measure yields mixed results on the linkages between investment
and their uncertainty measure. Ghosal and Loungani (1996) report a negative role of output uncertainty on investment.
 Exact

Leahy and Whited (1996),
 Suffix

using risk measures constructed from stock return data, argue that uncertainty exerts a strong negative effect on investment and point out that uncertainty affects investment directly rather than through covariances.
 (check this in PDF content)

 Start

10244
 Prefix

Leahy and Whited (1996), using risk measures constructed from stock return data, argue that uncertainty exerts a strong negative effect on investment and point out that uncertainty affects investment directly rather than through covariances.
 Exact

Guiso and Parigi (1999)
 Suffix

investigate the impact of demand uncertainty using firm
level data to show that uncertainty weakens the response to demand and
slows down capital accumulation. Minton and Schrand (1999) find evidence
that cash flow volatility is costly and leads to lower levels of investment in
5See Edmiston (2004) for other studies that concentrate on the linkages between investment and volatility in taxes.
6Some
 (check this in PDF content)

 Start

10425
 Prefix

risk measures constructed from stock return data, argue that uncertainty exerts a strong negative effect on investment and point out that uncertainty affects investment directly rather than through covariances. Guiso and Parigi (1999) investigate the impact of demand uncertainty using firm
level data to show that uncertainty weakens the response to demand and
slows down capital accumulation.
 Exact

Minton and Schrand (1999)
 Suffix

find evidence
that cash flow volatility is costly and leads to lower levels of investment in
5See Edmiston (2004) for other studies that concentrate on the linkages between investment and volatility in taxes.
6Some researchers have studied the extent to which a proxy for analysts’ forecasts can
explain firms’ investment behavior; see among others Abel and Eberly (2002).
5
capital expenditures, R&D
 (check this in PDF content)

 Start

10547
 Prefix

Guiso and Parigi (1999) investigate the impact of demand uncertainty using firm
level data to show that uncertainty weakens the response to demand and
slows down capital accumulation. Minton and Schrand (1999) find evidence
that cash flow volatility is costly and leads to lower levels of investment in
5See
 Exact

Edmiston (2004)
 Suffix

for other studies that concentrate on the linkages between investment and volatility in taxes.
6Some researchers have studied the extent to which a proxy for analysts’ forecasts can
explain firms’ investment behavior; see among others Abel and Eberly (2002).
5
capital expenditures, R&D and advertising.
 (check this in PDF content)

 Start

10797
 Prefix

Minton and Schrand (1999) find evidence
that cash flow volatility is costly and leads to lower levels of investment in
5See Edmiston (2004) for other studies that concentrate on the linkages between investment and volatility in taxes.
6Some researchers have studied the extent to which a proxy for analysts’ forecasts can
explain firms’ investment behavior; see among others
 Exact

Abel and Eberly (2002).
 Suffix

5
capital expenditures, R&D and advertising. Beaudry, Caglayan and Schiantarelli (2001) show that macroeconomic uncertainty captured through inflation variability has a significant effect on investment behavior of firms.
 (check this in PDF content)

 Start

11042
 Prefix

in taxes.
6Some researchers have studied the extent to which a proxy for analysts’ forecasts can
explain firms’ investment behavior; see among others Abel and Eberly (2002). 5
capital expenditures, R&D and advertising. Beaudry, Caglayan and Schiantarelli (2001) show that macroeconomic uncertainty captured through inflation variability has a significant effect on investment behavior of firms.
 Exact

Bloom et al. (2007)
 Suffix

suggest that higher uncertainty renders firms more cautious and reduces the effects of demand shocks on investment. Boyle and
Guthrie (2003) argue that offsetting effects of payoff and financing uncertainty must be distinguished in order to accurately gauge their effects on
investment.
 (check this in PDF content)

 Start

11178
 Prefix

Beaudry, Caglayan and Schiantarelli (2001) show that macroeconomic uncertainty captured through inflation variability has a significant effect on investment behavior of firms.
Bloom et al. (2007) suggest that higher uncertainty renders firms more cautious and reduces the effects of demand shocks on investment.
 Exact

Boyle and Guthrie (2003)
 Suffix

argue that offsetting effects of payoff and financing uncertainty must be distinguished in order to accurately gauge their effects on
investment.
Although these studies summarized above have examined various aspects
of the linkages between uncertainty and investment, none of them have entertained the impact of intrinsic or extrinsic uncertainty and aCAPMbased
risk measure in a regression model.
 (check this in PDF content)

 Start

12381
 Prefix

the analytical model used to link uncertainty faced by the firm to its choice of an optimal investment plan as well
as the method that we use to obtain our proxies for uncertainty.
3 An extendedQmodel of firm value optimization
The theoretical model proposed in this paper is based on the firm value
optimization problem and represents a generalization of the standardQ
models of investment by
 Exact

Blundell, Bond, Devereux and Schiantarelli (1992).
 Suffix

The present value of the firm is equated to the expected discounted stream
ofDt, dividends paid to shareholders, where 0< ρ <1 is the constant
oneperiod discount factor:
Vt= maxEt
[∞
∑
s=0
ρsDt+s
]
.(1)
At timet, all present values are known with certainty while all future variables are stochastic.
 (check this in PDF content)

 Start

14341
 Prefix

proxies of
financial health.7Finally, the firm faces the transversality condition which
prevents the firm from borrowing an infinite amount and paying it out as
dividends:
lim
T→∞
T−1∏
j=t
ρj
BT= 0,∀t.(5)
The first order conditions of this maximization problem for investment, capital and debt are
∂Ct
∂It
+ 1 =λt,(6)
∂Πt
−
∂Ct
∂Kt
=λt−(1−δ)ρEtλt+1,(7)
∂Kt
Et[ρRt+1] = 1 +μt,(8)
7See
 Exact

Hubbard, Kashyap and Whited (1995)
 Suffix

for a similar modeling strategy.
7
where the Lagrange multipliersλtandμtrepresent the shadow prices associated with the capital accumulation and the borrowing constraint, respectively. Equation (6) sets the marginal cost associated with an additional
unit of investment equal to its shadow price.
 (check this in PDF content)

 Start

15976
 Prefix

This term equals zero when the shadow price of external finance is equal to zero,μt=μt+i= 0∀i. We define averageQas
Qt=Vt/Kt−1and the leverage ratio asBt/Kt−1. For the unlevered firm
marginal q is equal to average Q in the case of no borrowing constraint.8
Similar to
 Exact

Love (2003),
 Suffix

we assume that adjustment costs are quadratic
and take the form
C(It, Kt, ε) =
b
2
[(
It
Kt
)
−g
(
It−1
Kt−1
)
−a+εt
]2
Kt.(11)
To obtain an investment equation, we rewrite the first order condition (6)
making use of the functional form of adjustment costs:
It
Kt
=a−
1
b
+g
It−1
Kt−1
+
1
b(1−δ)
Qt−
Rt
b(1−δ)
Bt
Kt−1
−
1
b
Θt
Kt−1
.(12)
8Hennessy (2004) obtains a similar result in which averageQove
 (check this in PDF content)

 Start

16287
 Prefix

case of no borrowing constraint.8
Similar to Love (2003), we assume that adjustment costs are quadratic
and take the form
C(It, Kt, ε) =
b
2
[(
It
Kt
)
−g
(
It−1
Kt−1
)
−a+εt
]2
Kt.(11)
To obtain an investment equation, we rewrite the first order condition (6)
making use of the functional form of adjustment costs:
It
Kt
=a−
1
b
+g
It−1
Kt−1
+
1
b(1−δ)
Qt−
Rt
b(1−δ)
Bt
Kt−1
−
1
b
Θt
Kt−1
.(12)
8
 Exact

Hennessy (2004)
 Suffix

obtains a similar result in which averageQoverstates marginalqby
incorporating postdefault returns to investment.
8
The last term in Equation (12) captures the role of financial frictions in the
firm’s capital investment behavior.
 (check this in PDF content)

 Start

18485
 Prefix

of
these sources would be operational when lenders evaluate the firm’s creditworthiness, we believe that use of firmspecific daily returns can provide
us with a single proxy which embodies all potential sources of uncertainty
relevant to the firm. Furthermore, using intrinsic and extrinsic uncertainty
in our regressions we can determine whether investment behavior is more
9See, for example,
 Exact

Almeida, Campello and Weisbach (2004).
 Suffix

10For instance, see Bloom et al. (2007) for a discussion of similar issues.
11Θt/Kt−1is a function of uncertainty and liquidity, evaluated as the ratio of cash flow
to the lagged capital stock. We apply a firstorder Taylor expansion to those factors to
derive this expression.
12We explain how these measures are constructed using daily data in section 3.1.
9
sensitive to own or marketspecific un
 (check this in PDF content)

 Start

18543
 Prefix

Furthermore, using intrinsic and extrinsic uncertainty
in our regressions we can determine whether investment behavior is more
9See, for example, Almeida, Campello and Weisbach (2004). 10For instance, see
 Exact

Bloom et al. (2007)
 Suffix

for a discussion of similar issues.
11Θt/Kt−1is a function of uncertainty and liquidity, evaluated as the ratio of cash flow
to the lagged capital stock. We apply a firstorder Taylor expansion to those factors to
derive this expression.
12We explain how these measures are constructed using daily data in section 3.1.
9
sensitive to own or marketspecific uncertainty.
 (check this in PDF content)

 Start

20090
 Prefix

The beginning of period averageQis defined as the market value
of the firm (shares plus debt) net of the value of current assets (inventories
and financial assets) divided by the replacement value of the firm’s capital
stock, imputed by the method of
 Exact

Salinger and Summers (1983).
 Suffix

13Finally,
Iis investment,CFdenotes cash flow andBis the firm’s debt. All terms
are deflated by the consumer price index taking into account the timing of
the variables appearing in the numerator and denominator.
 (check this in PDF content)

 Start

20764
 Prefix

We also anticipate that the cash flow sensitivity
of investment should increase in the presence of heightened uncertainty as
captured through the interaction terms.
13This methodology was also employed by
 Exact

Leahy and Whited (1996).
 Suffix

10
To summarize, our model contains three of the basic elements,Q, cash
flow and leverage, which have been shown to explain the investment behavior
of firms, along with three different measures of uncertainty.
 (check this in PDF content)

 Start

21445
 Prefix

, by themselves and interacted
with a measure of the firm’s liquidity, we can determine whether investment
behavior is more sensitive toOwn orM arketspecific uncertainty while the
covariance term helps us evaluate the predictions arising from theCAPM.
The interaction terms in the model allow us to examine whether uncertainty
makes managers more cautious in their investment decisions as
 Exact

Bloom et al. (2007)
 Suffix

claim.
3.1 Generating volatility measures from daily data
Any attempt to evaluate the effects of uncertainty on the firm investment
behavior requires specification of a measure of risk. The empirical literature
offers a number of competing approaches for the construction of volatility
measures.
 (check this in PDF content)

 Start

22829
 Prefix

Furthermore, a proxy for uncertainty
obtained from aGARCHspecification will be dependent on the choice of the
model and exhibit significant variation over alternatives.
In this study, we utilize daily stock returns and market index returns to
compute intrinsic and extrinsic uncertainty via a method based on
 Exact

Merton (1980)
 Suffix

from the intraannual variations in stock returns and aggregate financial market series.14This approach provides a more representative measure
14See Baum, Caglayan and Ozkan (2004) for a more detailed discussion of the procedure
11
of the perceived volatility while avoiding potential problems raised above.
 (check this in PDF content)

 Start

22991
 Prefix

In this study, we utilize daily stock returns and market index returns to
compute intrinsic and extrinsic uncertainty via a method based on Merton (1980) from the intraannual variations in stock returns and aggregate financial market series.14This approach provides a more representative measure
14See
 Exact

Baum, Caglayan and Ozkan (2004)
 Suffix

for a more detailed discussion of the procedure
11
of the perceived volatility while avoiding potential problems raised above.
Also the use of daily returns on the stock provides one with a forwardlooking
proxy for the volatility of the firms’ environment.
 (check this in PDF content)

 Start

24129
 Prefix

If data were generated every calendar day,
∆φt= 1,∀t,but given that data are not available on weekends and holidays,
∆φt∈(1,5).The estimated annual volatility of the return series is defined
as Φt[xt] =
√∑
T
t=1ς
d
twhere the time index for Φt[xt] is at the annual
frequency.
An alternative to Merton’s procedure (which makes use of squared highfrequency returns) is that proposed by
 Exact

Ghysels, SantaClara and Valkanov (2006)
 Suffix

: the computation of realized absolute variation and bipower variation, which make use of absolute returns. We generate these measures from
the firmlevel daily data, and find that when aggregated to the annual frequency they were correlated above 0.93 with our Mertonbased proxy.
 (check this in PDF content)

 Start

25070
 Prefix

on the S&P 500 index, inclusive of dividends.
4 Empirical findings
4.1 Data
The estimation sample consists of an unbalanced panel of manufacturing
firms for the 1984 to 2003 period drawn from Standard and Poor’s Industrial
along with its merits.
12
AnnualCOMPUSTATdatabase.15There are 9,895 firmyears for which the
replacement value of the real capital stock may be imputed by the method
of
 Exact

Salinger and Summers (1983).
 Suffix

A number of sample selection criteria are
then applied. We only consider firms which have not undergone substantial
changes in their composition during the sample period (e.g., participation
in a merger, acquisition or substantial divestment).
 (check this in PDF content)

 Start

27352
 Prefix

The screened data are used to reduce the potential impact of outliers upon the parameter
estimates.
13
4.2 The link between uncertainty and capital investment
In what follows we present our results in Table 2 obtained using the dynamic
panel data (DPD) approach developed by
 Exact

Arellano and Bond (1991),
 Suffix

as
implemented in Stata by Roodman (2007). All models are estimated in first
difference terms to eliminate unobserved heterogeneity using the onestep
GMMestimator on an unbalanced panel from a sample including 402 firms’
annual data.
 (check this in PDF content)

 Start

27404
 Prefix

The screened data are used to reduce the potential impact of outliers upon the parameter
estimates.
13
4.2 The link between uncertainty and capital investment
In what follows we present our results in Table 2 obtained using the dynamic
panel data (DPD) approach developed by Arellano and Bond (1991), as
implemented in Stata by
 Exact

Roodman (2007).
 Suffix

All models are estimated in first
difference terms to eliminate unobserved heterogeneity using the onestep
GMMestimator on an unbalanced panel from a sample including 402 firms’
annual data.
In column one, we start our investigation estimating a standard investment model which includes the basic explanatory variables for firm level
investment (Q,CFt/Kt−1andBt−1/Kt−2) along with the lagged depende
 (check this in PDF content)

 Start

28563
 Prefix

use of suitably lagged
endogenous variables as instruments.17In this model, neitherQnor the
cash flow/assets ratio appear significant, but the lagged measure of intrinsic
uncertainty has a negative and highly significant coefficient, while the interaction of the cash flow ratio with intrinsic uncertainty possesses a positive,
significant coefficient estimate. This is an interesting finding as
 Exact

Leahy and Whited (1996)
 Suffix

report that uncertainty affects investment behavior through
Q(in their analysis the coefficient on their proxy for uncertainty becomes
insignificant with the introduction ofQ). In our case, even in the presence ofQ(although insignificant), intrinsic uncertainty significantly affects
the firm’s fixed investment behavior, not only on its own but through the
interaction with cash flow.18
17The secon
 (check this in PDF content)

 Start

29206
 Prefix

), intrinsic uncertainty significantly affects
the firm’s fixed investment behavior, not only on its own but through the
interaction with cash flow.18
17The second through fourth lags of (It−1/Kt−2),Qt, (CFt/Kt−1), (Bt/Kt−1) and
lagged uncertainty measures are employed asGMMinstruments.
18The sceptical reader may be concerned about the possibility that Q may be measured
with error (See
 Exact

Erickson and Whited (2000)). To
 Suffix

dispel concerns that the use of lagged
variables as instruments may not overcome an issue of Q mismeasurement, we tested the
model by replacing Q with the ratio of (I/Kt+1+I/Kt+2)/(2I/Kt), following the approach
taken in Almeida et al. (2004).
 (check this in PDF content)

 Start

29456
 Prefix

To dispel concerns that the use of lagged
variables as instruments may not overcome an issue of Q mismeasurement, we tested the
model by replacing Q with the ratio of (I/Kt+1+I/Kt+2)/(2I/Kt), following the approach
taken in
 Exact

Almeida et al. (2004).
 Suffix

With that change made, the qualitative findings of Table 2
are virtually unchanged. The uncertainty proxies and interactions retain their significance,
even though this “perfect foresight” approach yields highly significant coefficients on the
alternative measure.
 (check this in PDF content)

 Start

31694
 Prefix

In contrast,extrinsic
uncertainty has a significant negative coefficient on the interaction term:
an increase in marketbased uncertainty decreases the incentive to invest at
any level of cash flow. Perhaps this finding suggests cautious behavior of
managers as in
 Exact

Bloom et al. (2007)
 Suffix

when uncertainty increases. Finally, we
observe that both the main effect and the indirect effect through theCAPMbased uncertainty term are significant. While the direct effect is negative,
the indirect effect is positive.
 (check this in PDF content)

 Start

32377
 Prefix

Taking all effects into account, the firm’s rate of investment
becomes more sensitive to available cash flow with an increase in uncertainty.
Q is not seriously biasing our findings.
15
This result, supporting the implications ofCAPMtheory, is quite interesting
and stands in clear contrast to the findings reported by
 Exact

Leahy and Whited (1996)
 Suffix

(although their model did not incorporate an interaction with cash
flow). In the model of column five, the cash flow ratio and debt ratio play
important roles in conjunction with uncertainty while Tobin’sQis generally
insignificant.
 (check this in PDF content)

 Start

33017
 Prefix

The overall
cash flow sensitivity of investment is an increasing function of uncertainty except forMarket(extrinsic) uncertainty. WhenMarketuncertainty increases,
the impact of cash flow is reduced. This can be explained in the light of
the findings of
 Exact

Beaudry et al. (2001),
 Suffix

who show that an increase in macroeconomic uncertainty would lead to a negative impact on firms’ investment
behavior as firm managers will not be able to readily distinguish good from
bad investment projects.
 (check this in PDF content)

 Start

34099
 Prefix

Even a casual inspection of these derivatives
shows that the role of uncertainty on firm investment is not trivial, and
varies considerably across types of uncertainty, in line with arguments put
forth by
 Exact

Boyle and Guthrie (2003).
 Suffix

20In particular, one can see that an
increase in the firmlevel (η) uncertainty measure leads to an increase in
19Tables of numerical values underlying the graphs are available from the authors upon
request.
20“. . . any attempt to empirically identify the relationship between uncertainty and investment will pick up offsetting uncertainty effects unless the exact nature of the uncertainty is caref
 (check this in PDF content)

 Start

37251
 Prefix

We specifically concentrate on the
role of firmspecific (intrinsic), marketspecific (extrinsic) andCAPMbased
measures of uncertainty on firms’ investment spending in that relationship,
allowing for interactions with cash flow. Both idiosyncratic and market uncertainty measures are constructed using a method based on
 Exact

Merton (1980)
 Suffix

from the intraannual variations in stock returns using firm level stock prices
and S&P 500 index returns. Employing annual data obtained fromCOMPUSTATfor manufacturing firms over the period between 1984–2003 we then
investigate the linkages between investment, cash flow and uncertainty.
 (check this in PDF content)