ECONOMIC RECORD, VOL.

90, SPECIAL ISSUE, JUNE, 2014, 87101

Uncertainty and Investment: Evidence from Australian

Firm Panel Data
THANG LONG TRAN
Department of Economics, Monash University, Clayton, Victoria, Australia,
National Economics University, Hanoi, Vietnam

This paper investigates the key drivers of fixed firm investment of

listed non-financial companies in Australia over the period from
1987 to 2009. A Tobins q model of investment is augmented to
account for the effect of economic uncertainty on the investment
decision. The effects of Tobins q, sales and cash flows on firm
investment rate are also analysed and discussed. Consistent with
existing literature, this research finds clear evidence of negative
effects of both macroeconomic and firm idiosyncratic uncertainty on
Australian firm investment. However, evidence also shows that firmspecific uncertainty is more important in explaining firm investment
than macroeconomic uncertainty.
parameters such as firms attitude towards risk,
the cost function, market competitiveness and the
shape of the marginal productivity of capital (see
Pindyck, 1982; Abel, 1983). For example, under
the conditions of perfect competition, an increase
in uncertainty stimulates investment if the marginal product of capital is a convex function of
wages and output prices, whose evolution is
uncertain. However, an increase in uncertainty
may discourage investment if the marginal product of capital function is concave.
At the firm level, Dixit and Pindyck (1994)
show that when there is uncertainty and irreversibility, the behaviour of firms is different. This is
because investment cannot be reverted after being
implemented and therefore firms need to account
for the uncertainty before they go ahead with the
investment decision. The option to delay the
investment decision is valuable, as firms can
decide whether to invest immediately or to wait
for new information. Greenwald and Stiglitz
(1993) argue that risk-aversion would make firms
want to invest less, whereas Bernanke (1983)
shows that firms postpone investing due to the
benefits that arise from the arrival of new information. Furthermore, Bernanke (1983) argues that
from the firms point of view, when investment
projects are even partly irreversible, high uncer-

I Introduction
The recent persistent decline in Australian
investment activity in the wake of the Global
Financial Crisis has reinforced the need to better
understand the factors that drive private business
investment, especially uncertainty. This paper
aims to revisit the topic of Australian firm
investment by undertaking a rigorous examination of key factors (dictated by their theoretical
significance) thought to impact firm investment
activity. Most importantly, this is the first paper
that examines the effect of uncertainty on
Australian firms investment.
While theoretical considerations appear to
support the conjecture that uncertainty is related
to firm investment, the sign and magnitude of the
relationship are not explained in a satisfactory
manner in both theoretical and empirical literature. From a theoretical point of view, the effect of
uncertainty on firm investment is ambiguous and
dependent on the relationships amongst the variables as well as on the assumptions of the model
JEL classifications: D80, D92, E22, G31
Correspondence: Thang Long Tran, Department of
Economics, Monash University, Clayton, Vic. 3800,
Australia.Emails: [email protected]; longtt@
neu.edu.vn

87
2014 Economic Society of Australia
doi: 10.1111/1475-4932.12133

88

ECONOMIC RECORD

tainty induces firms to delay investment decisions.

Moreover, uncertainty can depress investment
because of the increasing cost of finance, raising
managerial risk-aversion (Panousi & Papanikolaou, 2012), or an increase in agency problems
(DeMarzo & Sannikov, 2006).
The importance of two types of uncertainty,
idiosyncratic firm-specific uncertainty relative to
aggregate uncertainty, however, is not conclusive.
Caballero and Pindyck (1996) claim that aggregate uncertainty is more important than idiosyncratic uncertainty, whereas Leahy and Whited
(1996) find that aggregate uncertainty is not
significant and conclude that uncertainty affects
investment mainly through q for a panel of US
manufacturing firms. In contrast, Dixit and Pindyck (1994) argue that idiosyncratic uncertainty is
as important as aggregate uncertainty. In addition,
Bo (2002), using panel data of Dutch firms, finds
idiosyncratic uncertainty more important than
aggregate uncertainty in driving firm investment.
The majority of studies that examine the
investmentuncertainty nexus at the firm level,
focus on US and UK data (see for instance, Leahy
and Whited, 1996; Baum et al., 2008 and Bloom
et al., 2007), whereas only a handful of studies
investigate non-Anglo countries. 1 In contrast to
the research pertaining to the US economy where
both theoretical and empirical studies with
respect to investment have continued to give
insights into the behaviour of US firm investment
activity, the research investigating the determinants of Australian private firm investment since
the 1980s is extremely limited. For example,
Mills et al. (1995), La Cava (2005) and Chang
et al. (2007) are the only studies that analyse the
drivers of firm investment in Australia; however,
their findings are mainly on financial determinants or subject to specific caveats or objectives
when applied to Australian data. For instance,
Mills et al. (1995), using a small sample of 66
listed companies during an 11-year period, find
that certain financial factors had a significant
impact on investment decisions of Australian
companies. Their study is subject to small sample
size and possible selection bias. Chang et al.
(2007) investigate the impact of financial con1

See Guiso and Parigi (1999) and Bontempi et al.

(2010) examine Italian manufacturing firms; Fuss and
Vermeulen (2008) Belgium firms; Von Kalckreuth
(2003) German firms; Bo (2002) Dutch firms; Hatakeda
(2002) Japanese firms; and Pattillo (1998) examines
firms in Ghana.

JUNE

straints on Australian listed companies investment decisions and demand for liquidity over the
19902003 time period. They find that financial
constraints reduce the sensitivity of investment to
the availability of internal funds. To our knowledge, there is no study that examines uncertainty
and its effect on firm investment in Australia
where its main focus is something other than
financial determinants.
This paper is motivated by the need to investigate the role of uncertainty on the dynamics of
Australian firms investment. As a larger proportion of Australian listed firms are in the resource
sectors, thereby having more tangible assets, they
are more likely to be transparent and less subject
to market imperfections (Chang et al., 2007) and
their investments may be irreversible and large.
As such, it is important to investigate the impact
of uncertainty on Australian firm investment as
fundamental differences may exist between the
Australian and other countries listed firms,
which may lead to different outcomes of uncertainty.
This research contributes to the existing literature as new uncertainty proxies are incorporated
into the model using panel data of Australian
firms from 1987 to 2009, the largest time period
available. Although the primary measure of
uncertainty in our study is the volatility of returns
of firms stock prices, the paper additionally uses
new methods of measuring uncertainty. We specifically consider the effects of two different
forms of uncertainty on firms investment, which
include firm idiosyncratic (micro) uncertainty,
derived from either residuals obtained from a
conditional return model, or covariance between
the firm stock returns and market returns, and
market (macro) uncertainty measured based on
either AllOrds index returns or a GARCH model
of leading macroeconomic indicators.
The results of this study suggest that a negative
relationship exists between investment and uncertainty, while its effects depend on the different
proxies used and the nature of the firm. The other
explanatory variables, used by financial literature
including Tobins q, cash flow, leverage and
sales, are also found to be important to firms
investment. The coefficient of the variable measuring uncertainty is consistent and significant
with alternative models. The sign and strength of
the relationship depend upon the market power of
the firm and the degree of financial constraints.
The impact of uncertainty varies with firm size;
and, after controlling for fundamental vari 2014 Economic Society of Australia

2014

FIRM INVESTMENT AND UNCERTAINTY: AUSTRALIAN EVIDENCE

89

ables, firm-specific uncertainty, rather than

macroeconomic uncertainty, is more relevant for
investment decisions.
The next section discusses the empirical models to be utilised. Data sources and methodologies
are described in Section III. The empirical results
are discussed in Section IV, while the robustness
analysis is given in Section V. The final section
concludes the paper.

function of uncertainty, the discount rate and the

expected growth in the return to the asset. The
higher the uncertainty, the higher the threshold
value, the more likely the firm would delay its
investment.
Therefore, the q-based investment model
should be augmented with uncertainty, yielding:

II Empirical Models
Our empirical model inspired by Carrington
and Tran (2012) is a simplified investment model
augmented from the Tobins q theory. The
standard q model assumes that perfect foresight
and investment are solely decided by q, the ratio
between expected marginal revenue and the
marginal cost of additional unit of capital. The
starting model could be:

where Var is measure of uncertainty.

According to q theory, if the market is frictionless, q is a sufficient proxy for all firms
business opportunities. However, in reality, firm
investments are influenced by other factors; a
detailed discussion on these factors can be found
in Blanchard et al. (1993); Mills et al. (1995);
Romer and Romer (2000) and Romer (2006),
Ch. 8. Therefore, in our empirical model, together
with q and uncertainty, other factors generally
found to be important for investment are also
incorporated profitability, cash flows, leverage,
and sales. They have been incorporated to allow
for the possibility of credit constraints, divergence between the market and fundamental value
of capital, and the cost of uncertainty impacting
on investment.
Accounting for persistence of investment, lags
of explanatory variables are included. In addition,
the lag of dependent variable is included. It is
plausible that the effects of investment are
persistent and that one of the factors that affect
investment today is investments in the previous
periods. The empirical model to be estimated is
given by Equation (4):

I=Kt a0 a1 qt et

where I/K is investment ratio, K is capital stock,

I is investment, q t is Tobins q or average q, the
ratio between market value of capital and its
replacement costs, and it is expected that a 1 > 0.
Following Dixit and Pindyck (1994), when
there is uncertainty, firms may wait before
making investment decisions. There are similarities between an investment decision and a call
option (that is, invest now or wait later). If
uncertainty increases, the value of the option to
delay the investment can be substantial. Firms
will only invest now if the return of the investment exceeds the return of the delayed investment
plus the option value to wait.
Using some boundary conditions, Dixit and
Pindyck (1994, p. 146) show that the threshold
value of Tobins q at which the firm should invest
under uncertainty is given by:
qu

b
[1
b1

where b 1=2  r  d=r2

q ,
r  d=r2  1=22 2r=r2
r is the risk-free interest rate, d is a parameter and
r2 is the variance of share returns.
The threshold value which q has to exceed to
make new investment profitable now exceeds the
conventional value of 1 and is an increasing

2014 Economic Society of Australia

I=Kt a0 a1 qt d1 Var t et

Xk
I=Kt i;t a0 a1 I=Ki;t1
bq
j0 j i;tj
Xk
X
k

d Vari;tj
cZ
j0 j
j0 j i;tj
firmi year t mi;t

Because I/K can be negative (or disinvestment),

natural logs are not used in the model. Capital
stock is measured in this study as book value of
firms total assets. Tobins q is the ratio between
total market value of the firm and its replacement
costs, which is measured as market value of firms
equity plus book value of firms debt (calculated
as total assets minus common equity) scaled by
book value of total assets. The inclusion of lagged
q values is based on the notion that there are
delays between the date when the decision is
made to invest and the actual date when the

90

ECONOMIC RECORD

investment occurs. Var is uncertainty, measured

as the realised standard deviation of the stock
daily returns of the firm within the firm year. Z is
a vector of other controlling variables. Vector Z
includes firm sale ratio and the leverage ratio as
they have been highlighted as important determinants of firm investment in existing literature.
The details of how to measure these variables are
shown in Data section of this paper.
In the regression model, time dummies are
included to account for possible business cycle
effects as well as other macroeconomic factors
not mentioned in the model that influence all
firms from year to year. Firm dummies in the
fixed effect (FE) regression model are added to
account for other unobservable factors, which are
not proxied by explanatory variables and are time
invariant.
The baseline model should also be extended
with a firm size dummy that categorises firms by
their size. Financial literature claims that firm
size is a proxy for financial constraints and
information asymmetry between borrowers and
lenders, the differences in information available
to different parties in a financial contract (Fazzari
et al., 1988). The baseline model is extended with
lags of explanatory variables as contemporary
investment may be driven by past information and
also to avoid endogeneity issues. The use of
lagged uncertainty measures is motivated by the
fact that actual investment at time t is affected by
observed uncertainty in the environment in the
past periods.
The baseline model will be estimated for
comparison purposes using the pooled OLS,
HuberWhite variance/covariance matrix, and
generalised method of moments (GMM) estimators. Pooled regressions combine both dimensions
in one dataset by neglecting time and crosssectional structure, and heterogeneity across
subjects of interest, while assuming orthogonality
between regressors and the residual. According to
Wooldridge (2002), pooling increases the sample
size, thereby obtaining more precise estimates
and test statistics with greater power. Huber
White variance/covariance matrix estimator is
used to allow the residuals to be correlated within
years and heteroscedasticity, while GMM
dynamic panel estimator is used to account for
unobserved heterogeneity and possible endogeneity of the regressors, which FE model estimates
may suffer.
Theoretically, when the capital stock is higher
than its optimal level, the firm can disinvest by

JUNE

selling its fixed asset/capital (Abel & Eberly,

1999). In reality, firms investment is irreversible,
as selling plant and equipment is difficult, and if
successful, they may have to sell at a price well
below the real value of the plant and equipment.
Sale of existing property plants and equipment is
not always a choice even in the case where the
firm wants to disinvest. The data show an
asymmetry in investment, as most of investment
is greater than or equal to zero, while negative
investment is minority (approximately 9 per
cent).
Taking into account the asymmetric nature of
investment, a logit model is estimated to evaluate
the effect of uncertainty on the firms decision to
investment. In the logit model, the dependent
variable is a bivariate dummy, which takes the
value of 1 if investment in the period is positive
and 0 otherwise. The model tests the determinants
of a firms decision to invest (rather than the
amount of investment). According to the real
option theory, when uncertainty increases, firms
either wait or invest. Investment projects will be
implemented only if their returns have exceeded a
trigger value that reflects the cost of extinguishing the firms call option. Uncertainty increase
will raise the trigger value, which delays investment, or reduces the probability of positive
investment. Hence, this theory is tested by
regressing a panel logit model based on the
following equation:
Xk
InvDumi;t a0 a1 InvDumi;t1
bj qi;tj
Xk
Xk j0

k Z
c Var i;tj
j0 j i;tj
j0 j
firmi year t mi;t

5
where InvDum is the binary variable of investment; taking the value of 1 if investment is
positive and 0 otherwise. The probability of
positive investment at time t is allowed to depend
on the outcome of investment in t  1 as well as
q, uncertainty and other variables.
III Data
Annual accounting data of all Australian listed
firms, available from 1987 to 2009, are obtained
from the Aspect Fin Analysis. These companies
are or were listed on the Australian Securities
Exchange (ASX). Data on firms daily stock
prices are from the Share Price and Price Relatives (SPPR) dataset from 2 January 1987 to 31

2014 Economic Society of Australia

2014

FIRM INVESTMENT AND UNCERTAINTY: AUSTRALIAN EVIDENCE

December 2009. The fiscal year for most firms is

from July 1st to June 30th of the following year.
Our firm population includes all listed as well as
delisted companies on the ASX so to avoid
survivor biases. The firms for the sample were
selected using a simple selection criterion. All
Australian firms were included in the sample
except for those firms in the financial, utility and
insurance and property sectors. Insurance, property and financial firms are excluded because of
their relatively low physical capital investment,
while utility firms are excluded due to their
regulated nature. The following couple of paragraphs discuss adjustments to the dataset. The end
result of the adjustments is an unbalanced panel
dataset of 1,235 companies over the 19872009
period, with total 12,175 firm years.
To control for outliers, the growth rate of each
firms total assets is calculated, and the annual
distribution of this growth rate above the 98th
percentile is cut to remove firms exhibiting
substantial changes in firm size. To exclude firms
that show a clear evidence of financial distress,
the firms that have the cash flow ratio (calculated
as total profit after tax before abnormal items plus
depreciation scaled by beginning-of-period capital) of any given year below 50 per cent are
excluded. This measure is utilised as an exclusion
criteria because when the cash flow loss accounts
for more than 50 per cent of total assets, firms are
in financial distress.
As this study focuses more on middle and large
firms, those with capital lower than one million
AUD (year 2009 prices) in any year of their life
are excluded. The reason this study focuses more
on middle and large firms is to avoid any
financing frictions that might be present for
smaller firms and to avoid the bias in the model
resulting from small capital stock, which causes
values of variables to become abnormally large,
given the small denominators in the ratios. Later
on, when the minimum capital threshold is
changed to two and three million AUD, regression results are qualitatively similar.
The data are checked for large discontinuities in
the book value of capital that cannot be explained
by investment, capital expenditures or depreciation reported by the firm. If large discontinuities
are present, then those years/firms are removed.
Following Bond and Meghir (1994), only companies that appear for at least five consecutive years
in the data are kept in the sample. To account for
mergers or large acquisitions, the delisted firms
last years observations, with delist codes indi 2014 Economic Society of Australia

91

cating merging and acquisitions are dropped.

Furthermore, the 1st and the 99th percentile of
all main variables (q, investment, cash flow, sales
and leverage ratios) is winsorised. This is done by
setting the tail values equal to some specified
percentile of the data; here, they are the 1st and
the 99th percentile. Following Barnett and Lewis
(1994), this approach reduces the impact of
extreme observations by assigning a cut-off value,
thereby removing values past the cut-off point.
At firm level, uncertainty can be classified into
two types idiosyncratic and aggregate uncertainty. Idiosyncratic uncertainty is firm-specific
uncertainty and it is independent across firms.
Firms may be uncertain about two types of
factors, firm-specific factors, such as output
prices, demand, input costs, capital costs, technology, firm structure and organisation, and
future profitability; and factors related to the
external environment, such as, the movements of
exchange rates, inflation, interest rates or stock
markets. Measuring uncertainty for specific firms
is difficult because reasonably high-quality data
on the variables of interest (such as output prices,
input costs, demands) are not available on a
sufficiently disaggregated basis (Leahy &
Whited, 1996). In the existing literature on the
investmentuncertainty relationship, several
methods have been used for measuring uncertainty. First, one can use the unconditional
variance of the historical data of the variables
of interest as a proxy for uncertainty. The second
method measures uncertainty as the volatility of
the unpredictable part derived from a prediction
model (Koetse, 2006). Another popular method is
measuring uncertainty by a GARCH-based model
of volatility along the lines of Episcopos (1995).
In this paper, uncertainty is primarily measured
by the first method, using the within-year volatility of the firms daily stock returns, measured
as the standard deviation of stock returns during
the firms current accounting period. 2 This
2
Estimated annual volatility of the return series is
defined as average standard deviation of daily returns
within the respective year before the dates of financial
reports

s
P
x  x2 p
 250
Volatility
n
where x is daily return of stock price, n is the number of
trading days.

92

ECONOMIC RECORD

method is widely used by researchers, for example, see Caballero and Pindyck (1996), Leahy and
Whited (1996), Bloom et al. (2007) and Baum
et al. (2008). Stock prices are forward-looking
measures of profitability, containing both firm
uncertainty and market uncertainty. Other measures of uncertainty are discussed in the robustness section, which give qualitatively similar
results.
(i) Descriptive Statistics
Table 1 summarises the construction of the
variables used in the model. The descriptive
statistics of firms in the sample are presented in
Table 2. These results are comparable to the
result of Chang et al. (2007), who examine
investment and cash flow of listed Australian
firms over the period 19902003.
The term multicollinearity is used to describe
the situation when a high correlation is detected
between two or more predictor variables. To
check for multicollinearity, the correlation matrix
is presented in Table 3.
The correlation matrix gives some insights into
the possible relationships between the variables
of interest. The values obtained are similar to
previous research using Australian firms (for
example, Mills et al., 1995 and Chang et al.,
2007). We see, however, that the correlation
coefficients are not too large, which could lower
the possible bias of multicollinearity in the
regressions. The average Tobins q is 1.67, which
is larger than one. Investment is positively
correlated with Tobins q, and negatively correlated with leverage, uncertainty and cash flow.

JUNE

T ABLE 2
Descriptive Statistics of Main Variables
Variable

Mean

SD

Min

Max

Investment
Sales
Cash flow
Leverage
Tobins q

0.094
0.753
0.025
0.41
1.67

0.233
1.001
0.432
0.449
2.713

0.258
0
2.60
0.006
0.282

1.497
5.181
0.624
3.327
20.024

The negative correlation coefficient between

investment and cash flow is consistent with the
findings in La Cava (2005); however, this is
inconsistent with the intuition that corporate
investment depends positively on internal cash
flow, which, in many cases, implies the existence
of financial constraints. This counter-intuitive
correlation between cash flow and investment
will be discussed in the forthcoming section.
IV Empirical Results
In this section, the main empirical model is
estimated using different estimation methods to
analyse the role of uncertainty in investment by
controlling for various variables.
To answer the question what factors drive the
firms decision to invest? a logit model regression of the investment indicator InvDum i,t (taking
the value one if investment is positive and zero if
investment is zero) as in Equation (4) is run, on
the explanatory variables used in our baseline
model. Hence, the regression tests the probability
of investment being greater than zero, given the

T ABLE 1
Measurement of Variables
Variables
Investment
Capital
Investment ratio
Cash flow
Sales
Cash flow ratio
Sale ratio
Leverage
Tobins q
Total market value
of equity

Measurement method
Purchase of property plants and equipment (PPEs) Sales of PPEs
Total book assets
Investment/Capital at the beginning of the year
Profit after tax before abnormal + Depreciation and amortisation
Total revenue
Cash flow/Capital at the beginning of the year
Sales/Capital at the beginning of the year
Total debt/Total book value of equity
(Total market value of equityBook value of equity + Book value
of total asset)/Total asset
Share price at the year-end times total outstanding number of shares

2014 Economic Society of Australia

2014

93

FIRM INVESTMENT AND UNCERTAINTY: AUSTRALIAN EVIDENCE

T ABLE 3
Correlation Matrix

Investment
Cash flow
Sales
Tobins q
Leverage
Uncertainty

Investment

Cash flow

Sales

Tobins q

Leverage

Uncertainty

1.00
0.08*
0.06
0.15*
0.07*
0.06*

1.00
0.29
0.10*
0.07*
0.22*

1.00
0.04*
0.27*
0.13*

1.00
0.15*
0.05*

1.00
0.04*

1.00

Note: *Significance at 5 per cent level.

explanatory variables. The role of q, uncertainty

and other variables in the logit model is to predict
the probability of investment/disinvestment,
rather than the amount of investment/disinvestment. The logit regression results in column 1 of
Table 4 show that the investment decision is
driven by the anticipated variables (including
Tobins q, cash flow, sales and uncertainty). The
first lag of Tobins q plays a positive and
significant role on the investment decision.
Lagged sales and cash flow both play a positive
and significant role on the investment decision,
while lagged leverage plays a negative role; all
the results are as expected. The higher the debt is
as a ratio of total assets, the lower the probability
of the firm carrying out the investment project.
Most importantly, the results suggest that uncertainty plays a negative role in the investment
decision, and the option theory is relevant to
investment. Investment projects are implemented
because their returns have exceeded a trigger
value that reflects the cost of extinguishing the
firms call option. In short, given the above
results, it is clear that the investment decision is
driven by the information corresponding to the
firms business situation in the previous year (that
is, first lag of sales, first lag of leverage and
profitability) and the known business opportunities proxied by q.
The data are pooled to examine the possible
effects of control variables and uncertainty on
investment decisions. The results of the pooled
model presented in Table 4, column 2, confirm
the expected relationship between the investment
decision and its determinants. The coefficient of
uncertainty is statistically significant and negative, implying that investment is inversely related
to uncertainty. The coefficients of lagged Tobins
q and the leverage variables are of the correct
sign and are significant.
2014 Economic Society of Australia

The results of the baseline regression on panel

data considering the impact of a change in
Tobins q, cash flow, sales, uncertainty and the
interaction between sales and uncertainty on
investment are shown in columns 39 of
Table 4. Only the contemporaneous and the first
lag of explanatory variables are presented as the
longer lags are not significant in the regression.
This shows that the investment decision in time t
may be influenced only by the information in the
current period, t, and in the preceding one, t1.
The estimator used in columns 38 is the FE
estimator, as the Haussmann test results suggest
that FE specification is preferred to the randomeffect specification. The FE model contains firmspecific effects that are time-invariant and unobservable, but arbitrarily correlated with other
regressors, and is likely to be more appropriate in
this case as the sample consists of a broadly
exhaustive population (as differences between
individuals may be viewed as parametric shifts in
the regression function).
The results show that in the augmented model,
q t1 and sales are significant in terms of the
investment rate. Column 3 shows the results of
the static model, which excludes the lag of the
dependent variable, while in column 4 onwards,
the first lag of the dependent variable is included.
While column 4 is the baseline model regression,
the results from column 3 are almost the same
with those of column 4, showing that the inclusion of the lag of dependent variable does not
change the estimation result. Tobins q, sales,
leverage and uncertainty are the main variables
that affect investment; cash flow is only significant at the 10 per cent level. After controlling for
q, the proxy for business opportunity, uncertainty
has negative and significant effect on investment.
Given the sum of coefficients, it seems likely that
the effect of uncertainty is as large as that of

BIG 9
uncertainty t
sales 9
uncertainty t
constant

uncertainity t1

uncertainity t

leverage t1

leverage t

sales t1

sales t

cashflow t1

Cashflow t

q t1

qt

I/K t1

Dependent
variable

n/a

0.75***
(10.21)
0.04
(1.26)
0.12***
(3.05)
0.06
(0.19)
1.27***
(3.98)
0.45
(5.32)
0.06
(0.77)
0.21
(0.99)
0.38*
(1.71)
0.21***
(4.93)
0.09**
(2.27)

(1)
Logit (panel)
Investment
indicator

0.01

0.40***
(21.34)
0.01*
(1.77)
0.01***
(3.80)
0.02
(1.07)
0.00
(0.29)
0.03***
(5.10)
0.03***
(6.51)
0.06***
(3.38)
0.10***
(5.07)
0.01**
(2.08)
0.00
(1.45)

I/K t

(2)
Pooled
regression

0.05

0.01*
(1.78)
0.02***
(5.88)
0.04*
(1.78)
0.02
(1.16)
0.03***
(5.40)
0.01
(1.38)
0.07***
(3.23)
0.10***
(4.20)
0.01***
(4.21)
0.01
(1.65)

I/K t

(3)

0.05

0.13***
(6.14)
0.01*
(2.07)
0.02***
(5.67)
0.03
(1.29)
0.02
(1.05)
0.04***
(5.61)
0.01**
(2.40)
0.06***
(3.03)
0.10***
(4.28)
0.01***
(3.85)
0.00
(1.02)

I/K t

(4)

0.01

0.06**
(2.17)
0.00
(0.22)
0.03***
(4.64)
0.05
(1.04)
0.12***
(2.88)
0.02***
(3.19)
0.00
(0.10)
0.12***
(3.61)
0.17***
(5.20)
0.01*
(1.75)
0.00
(0.73)

I/K t

(5)
Large
firms

T ABLE 4
Results of the Baseline Model

0.12***

0.09***
(3.26)
0.01***
(2.99)
0.01***
(4.15)
0.00
(0.13)
0.01
(0.47)
0.04***
(4.04)
0.01**
(2.14)
0.02
(0.94)
0.05
(1.73)
0.01***
(3.49)
0.00
(0.90)

I/K t

(6)
Small
firms

0.05

0.13***
(6.08)
0.01**
(2.14)
0.02***
(5.66)
0.03
(1.24)
0.02
(1.12)
0.03***
(5.53)
0.01**
(2.41)
0.06***
(2.98)
0.10***
(4.23)
0.01***
(4.67)
0.00
(0.64)
0.01***
(3.09)

I/K t

(7)

0.01*
(1.75)
0.04

0.13***
(6.13)
0.01**
(2.06)
0.02***
(5.64)
0.03
(1.48)
0.02
(0.97)
0.02***
(3.06)
0.01**
(2.02)
0.07***
(3.92)
0.11***
(5.84)
0.02***
(4.31)
0.00
(1.00)

I/K t

(8)

0.19***
(6.79)
0.01*
(1.92)
0.01***
(3.87)
0.04
(0.70)
0.07**
(2.24)
0.04**
(2.32)
0.02**
(1.98)
0.02
(0.25)
0.01
(0.16)
0.02***
(3.55)
0.01
(1.44)

I/K t

(9)
Difference
GMM

94
ECONOMIC RECORD
JUNE

2014 Economic Society of Australia

 2014 Economic Society of Australia

6,208

(1.13)
0.25
10,254

I/K t

(2)
Pooled
regression

(1.33)
0.03
10,254

I/K t

(3)

(1.48)
0.10
10,254

I/K t

(4)

(0.69)
0.10
5,395

I/K t

(5)
Large
firms

(8.35)
0.09
4,859

I/K t

(6)
Small
firms

(1.39)
0.10
10,254

I/K t

(7)

(1.22)
0.14
10,254

I/K t

(8)

0.98

8,834
0.71

I/K t

(9)
Difference
GMM

Notes: *P < 0.1, **P < 0.05; ***P < 0.01. t statistics are in parentheses (z statistics are shown in column 1). q is Tobins q, sales is the ratio between total sale
revenue and beginning-of-period capital. cash flow is ratio between cash flow and beginning-of-period capital, leverage is ratio between total debt and total asset.
uncertainty is measured as the volatility of firms stock returns. Investment t is investment ratio. BIG is a binary variable proxying for the size of firms. The
pooled regression is estimated using OLS while standard errors are robust (Eicker Huber White) standard errors to account for heteroscedasticity and
correlation of disturbances within groups. The first regressor in column 1 is the first lag of dependent variable. Except column 1, all models include year dummies
to account for possible business cycle effects and unobservable variables that vary over time. The estimations in columns 2 8 correct the error structure for
heteroscedasticity and clustering using the White Huber estimator; hence standard errors are robust to arbitrary autocorrelation and heteroscedasticity.
Estimator in column 8 is one-step difference GMM. The instruments used in column (9) are the second to sixth lags of q, cash flow, sales, leverage, year dummies
and all the lagged measures of uncertainty. Instrument validity is tested using a Sargan Hansen test of the overidentifying restrictions. Second-order serial
correlation in the first-differenced residuals is tested using a Lagrange multiplier test (Arellano & Bond, 1991). The panel data are unbalanced.

R2
N
Hansen test
(P-value)
AR(2)

Dependent
variable

(1)
Logit (panel)
Investment
indicator

T ABLE 4
(continued)

2014
FIRM INVESTMENT AND UNCERTAINTY: AUSTRALIAN EVIDENCE

95

96

ECONOMIC RECORD

Tobins q. This result is consistent with other

studies using a similar q model for the US and
UK (for example, Bond et al., 2005 and Bulan,
2005).
Most of the estimated coefficients of Tobins q
are significant, and the magnitude is on average
0.02, which suggests that Tobins q has a moderate influence on investment, consistent with
previous studies on US firms. For instance, Leahy
and Whited (1996) find that the coefficient of
Tobins q ranges from 0.022 before controlling
for cash flow to 0.04 after controlling for cash
flow. Many other studies also find evidence of a
small coefficient of Tobins q; Bond et al. (2005)
find that the Tobins q coefficient for UK firm
investment is below 0.02; Mills et al. (1995) find
a Tobins q coefficient of 0.018 for 66 Australian
firms, whereas Baum et al. (2008) find that the
coefficient for Tobins q is also small in their
investment regressions.
Columns 5 and 6 contain results for two
different samples of firms that are based on firm
size. Using capital data, the firms are categorised
into two groups large and small. The coefficient
of lagged uncertainty on small firms is negative
and significant, while those on large counterparts
are not significant. This interesting result shows
that small firms are more vulnerable to uncertainty than large firms. This is a plausible result
because small firms are more sensitive towards
changes in the business environment compared
with large firms.
In column 7, a regression with an interaction
term between firm size and uncertainty is run, as
it is believed that large firms suffer less asymmetric information problems (between firms and
fund providers) and have fewer difficulties in
accessing external capital markets compared with
small firms (see, for example, Fazzari et al.,
1988). Firm size is defined by a big firm dummy
variable, BIG it, that takes the value 1 when total
assets are above the sample median within a
certain year, and 0 otherwise. This dummy is
interacted with the uncertainty variable to capture
any different effects uncertainty may have on
financially constrained firms, while naturally, the
coefficient on the uncertainty term by itself
captures the responsiveness of investment to
uncertainty for unconstrained firms. The coefficient on the interactive term is found to be
significant, which means that uncertainty plays a
much larger role in the investment decisions of
financially constrained firms compared with
unconstrained firms.

JUNE

Furthermore, column 8 regression includes an

interaction term between the uncertainty measure
and the sale ratio that is significant and positive,
indicating a stronger effect of sales on investment
at higher levels of uncertainty.
Unlike previous studies on firm investment in
Australia (La Cava, 2005 and Chang et al., 2007)
that find that cash flow positively and significantly affects investment, the results in this study
show that cash flow is not very significant. The
different results obtained could be a result of the
selection criteria of firms in the sample. For
example, Chang et al. (2007) eliminated all firms
with a minimum stock of property, plant and
equipment of less than $A5m, leaving them with
420 firms. However, La Cavas (2005) sample
consists of 300 firms and 1,700 observations over
the 19902004 period. Our sample, however,
contains 1,235 companies over the 19872009
period, a total of 12,175 observations.
Analogous to La Cava (2005), we find that
many of the listed/delisted firms on the ASX have
had negative cash flow over a large number of
consecutive years without being delisted from the
ASX. Numerous firms in industries such as
materials and business services (which are the
main industries in Australia) have had more than
60 per cent of their operating years exhibiting
negative cash flow. While firms that show clear
evidence of financial distress are excluded, nevertheless, there are still many firms that exhibit
negative cash flow during their years of operation.
As aforementioned, due to the irreversible nature
of investment, firms rarely sell property (especially plants and equipment) when they want to
disinvest. In general, the evidence shows that the
effects of cash flow on investment in Australia are
relatively insignificant. This is thought to be due
to the asymmetry between the distributions of
cash flow around zero, which causes us to find an
insignificant effect of cash flow on investment.
It is plausible that the effects of investment are
persistent and that one of the factors that affects
investment today is investments in the previous
periods; therefore, one lag of the dependent
variable is added to account for this notion. By
adding a lagged dependent variable, we encounter
endogeneity issues, which lead us to use the
difference GMM estimator when estimating the
dynamic panel model. GMM has been widely
used to deal with unobserved heterogeneity and
endogeneity bias in estimation.
The lags of the Tobins q, cash flow, leverage,
and sales, as well as lags of measures of uncer 2014 Economic Society of Australia

2014

FIRM INVESTMENT AND UNCERTAINTY: AUSTRALIAN EVIDENCE

tainty, are employed as GMM instruments (Arellano & Bond, 1991). By using GMM, which tests
for endogeneity, we take into account the endogeneity between investment and uncertainty. In
fact, firm-level uncertainty is not purely exogenous. For example, the decision to undertake a
risky investment project may introduce heightened uncertainty over the firms future returns.
Latent factors may also affect both uncertainty
and the attractiveness of investment, creating a
non-causal correlation. Following prior literature,
endogenous uncertainty has been handled by
using internal instruments: lagged values of
the dependent and explanatory variables (see, for
example, Leahy & Whited, 1996; Bulan, 2005 and
Bloom et al., 2007).
The results of the dynamic model are presented
in column 9 of Table 4 and support the results of
the q-based model. Tobins q, sales and uncertainty again are determinants of firm investment
with the coefficients having the consistent signs
and comparable magnitudes. The lag of the
dependent variable plays a moderately significant
role in deciding investment, showing the persistence of investment. The consistency in the
results amongst the different estimators confirms
the consistent behaviour of Australian firms
investment.
V Different Measures of Uncertainty
In this section, the relationship between investment and uncertainty is examined with some
alternative measures of uncertainty. Various
kinds of proxies have been used to measure
uncertainty in the literature. At the firm level,
uncertainty can be classified into macroeconomic
uncertainty, industry-wide uncertainty and idiosyncratic (or firm-specific) uncertainty. Macroeconomic uncertainty can be measured based on
the volatility of macroeconomic variables, such
as aggregate demand, exchange rates, interest
rates and inflation, just to name a few. To date, a
large amount of studies utilise macroeconomic
uncertainty because of the availability of macro
data, while only a small amount utilise firm-level
uncertainty. It is difficult to distinguish between
the industry and the firm idiosyncratic sources of
uncertainty in empirical research, mainly due to
data-related constraints; therefore, empirical
studies that distinguish explicitly between these
two sources of uncertainty are also scarce (Koetse, 2006).
The robustness of the role of uncertainty on
Australian firm investment is tested with the
2014 Economic Society of Australia

97

following changes to the uncertainty measures.

First, macroeconomic uncertainty is incorporated
and measured using two sources. The first measure is based on the volatility of All Ordinaries
Stock Index on the Australian Stock Exchange.
The use of the volatility of stock market returns
as a measurement of uncertainty has the advantage as, in principle, it captures all relevant
sources of risk. However, according to Guiso and
Parigi (1999), share prices may also respond to
extraneous information, reflect irrational behaviour and the presence of noise traders, or be
dominated by speculative bubbles and subsequent
crashes rather than by changes in the firms
fundamentals or in its perceived uncertainty. This
urges one to think about other proxies for
uncertainty that have been used in literature.
Another proxy for macroeconomic uncertainty
is constructed from the conditional standard
deviation of the index of leading economic
indicators 3 (here is Westpac Melbourne Economic Leading Index) as the index can be claimed
as a measure of overall macroeconomic activity.
The conditional variance of the index of leading
indicators is estimated with a generalised ARCH
(GARCH) model, where the mean equation is
autoregressive process AR(p), allowing for
ARMA errors. The twelve-month moving average
of the conditional standard deviation is computed.
Second, firm-specific uncertainty is measured
via several proxies that have only been used
recently by Bond et al. (2008). In existing literature on firm investment, two methods that have
been used for measuring firm uncertainty are that
one can take the unconditional standard deviation
of a series and use it as a proxy for uncertainty.
Another method is to use a more complicated
prediction model to take out the predictable part
of a time series, and then measure the uncertainty
by the volatility of the unpredictable part (Koetse,
2006). In this study, the proxy for firm-specific
uncertainty is the volatility of the unpredicted
part of the capital asset pricing model (CAPM),
3

Melbourne/Westpac leading index of economic

activity or Westpac Leading index is provided monthly
by Westpac Melbourne Institute Indexes of Economic
Activity reports. The data are from Datastream, beginning from January 1960. The estimated standard deviation used as uncertainty in the investment model is the
average of conditional standard deviation estimated
from a GARCH(1,1) model of the respective months.
The mean equation is autoregressive process order of
four AR(4). Our method is similar to Episcopos (1995).

98

ECONOMIC RECORD

similar to that in Baum et al. (2008). We start

from a conditional return model (such as CAPM),
where the stock return is conditionally predicted
based on the market stock return and beta that is
specific for the company. The unpredicted part of
the regression or the residuals is unobservable.
The standard deviation of the residuals can
therefore be used as a proxy for uncertainty.
To allow for time variation in the regression
coefficients and therefore, indirectly, possible
time variation in the factor loadings, rolling
regressions on the CAPM equation are
implemented to estimate firm-specific excess
returns. Daily returns are calculated as the first
differences in the natural logarithms of the
respective indices. To calculate daily excess
returns, the daily 90-day dealer bill rate is
subtracted and is available as a proxy for the
risk-free rate. Beta is then calculated using
CAPM to withdraw the residuals, and uncertainty
is therefore measured as the annual standard
deviation of the residuals series.
CAPM theory implies to measure uncertainty
by using the covariance between the firm stock
returns and market returns. However, as a result
of having no data on the risk of investment
projects in individual firms, it is assumed that risk
of an investment is equal to the risk of the firms
that carry out the investment. According to
CAPM, the required rate of return on an investment should be positively related to that investments risk, which is measured by the covariance
of its returns with the market as a whole. An
increase in the covariance should increase the
riskiness of investment, increasing the required
rate of return and reducing the desired level of
capital stock (Leahy & Whited, 1996). Using
covariance between stock returns and market
returns as proxy for uncertainty, Baum et al.
(2008) find a significant effect of uncertainty on
investment even in the presence of q, cash flow
and the debt-to-capital ratio. They consider this
covariance as the interaction between firm-specific and macroeconomic uncertainty. Therefore,
here we have two new measures of firm-specific
uncertainty based on the CAPM. First is firmspecific uncertainty obtained from the residuals,
and the second is the covariance between the firm
stock returns and market returns.
The one-step Difference GMM estimator is
used to run the regressions. The dynamic model is
estimated by including a lag of the dependent
variable extended with firm-specific and macroeconomic uncertainty. The results on different

JUNE

proxies for uncertainty are shown in Table 5. In

columns 1 and 2, uncertainty is macroeconomic
uncertainty, proxied by volatility of AllOrds
Stock Index returns and conditional standard
deviation of the Westpac Economic index
obtained by a GARCH model respectively. Columns 3 and 4 are models where firm-specific
uncertainty is included. In these columns, uncertainty is measured, respectively, as the unconditional standard deviation of firm share price
returns, the covariance between firm share price
returns and market stock index returns and the
standard deviation of residuals obtained from
CAPM-based model regressions for firms share
price returns on market portfolio returns. Both
contemporaneous and lagged proxies of macroeconomic uncertainty are used. In column 5, both
macroeconomic uncertainty and firm-specific
uncertainty are included. The J statistics (and
the corresponding p-value) is the HansenSargan
test statistic, and it indicates that the test for
overidentifying restrictions is satisfactory. Furthermore, we reject the presence of second-order
autocorrelation (AR(2)).
We see from Table 5 that regardless of the way
firm-specific uncertainty is measured, it consistently has a negative effect on firm investment.
We find no evidence that macroeconomic uncertainty measured by the volatility of the stock
market index (via AllOrds Stock Index) and
GARCH of the Westpac Melbourne Economic
Leading Index has a significant effect on firm
investment, especially when the model includes
firm-level uncertainty. This may be because
macroeconomic uncertainty affects firm investment via the volatility of firm shares, or because
investment decisions are influenced mainly by
firm idiosyncratic uncertainty. Firms could pay
more attention to its specific condition to consider
investment decisions. In addition, as time dummies are added to the models to proxy for the
unobservable time variant variables, they could
be highly correlated with business cycles, reducing the role of our proxy of macroeconomic
volatility. When the time dummies are taken out,
coefficients of the macroeconomic uncertainty
variables become significant.
Our results are consistent with empirical findings in other countries. Baum et al. (2008)
examine the uncertaintyinvestment relationship
for US firms and find that firm-specific and
CAPM-based uncertainty has a significant and
negative effect on investment. Bo (2002) provides
evidence that firms investment is more sensitive
2014 Economic Society of Australia

2014

99

FIRM INVESTMENT AND UNCERTAINTY: AUSTRALIAN EVIDENCE

T ABLE 5
Sensitivity Analysis for Different Measures of Uncertainty
Variables
Investment t1
qt
q t1
cashflow t
cashflow t1
sales t
sales t1
leverage t
leverage t1

(1)

(2)

(3)

(4)

(5)

(6)

0.20***
(7.01)
0.01
(1.55)
0.02***
(4.34)
0.09
(1.12)
0.07**
(2.01)
0.00
(0.11)
0.01
(0.73)
0.09**
(2.06)
0.09**
(2.15)

0.20***
(7.09)
0.00
(0.85)
0.02***
(3.85)
0.09
(1.33)
0.06*
(1.71)
0.01
(0.99)
0.01
(1.34)
0.04
(1.00)
0.08*
(1.87)

0.20***
(6.97)
0.01
(1.48)
0.02***
(4.34)
0.09
(1.18)
0.07**
(2.01)
0.00
(0.02)
0.01
(0.70)
0.09**
(2.12)
0.09**
(2.22)

0.19***
(6.96)
0.01*
(1.84)
0.02***
(4.48)
0.06
(0.94)
0.07**
(2.33)
0.02
(1.44)
0.01
(1.44)
0.04
(0.78)
0.06
(1.40)

0.18***
(6.63)
0.01*
(1.87)
0.01***
(3.92)
0.03
(0.61)
0.06**
(2.07)
0.04**
(2.49)
0.02**
(2.08)
0.01
(0.21)
0.02
(0.34)
0.02**
(2.72)
0.01
(1.26)
0.56
(1.63)
0.32
(1.04)

0.20***
(6.94)
0.01
(1.47)
0.02***
(4.24)
0.09
(1.14)
0.07**
(1.97)
0.00
(0.07)
0.01
(0.79)
0.10**
(2.20)
0.09**
(2.25)
0.01**
(2.59)
0.00
(0.18)
0.34
(1.22)
0.09
(0.24)
1.04***
(3.85)
2.96***
(5.01)

0.95
0.74

0.62
0.11

sdstockre t
sdstockre t1
sdallordsret
sdallordsret1

0.06
(0.34)
0.06
(0.29)

1.05**
(3.14)
2.24**
(2.84)

cov t
cov t1
sdresid t
sdresid t1
avsdwestpac t
avsdwestpac t1
AR(2)
Hansen J test P-value

0.70
0.11

2.25
(0.11)
7.09
(0.32)
0.67
0.13

0.65
0.10

0.03**
(2.54)
0.00
(0.27)

0.91
0.52

Notes: t-values are in parentheses, a full set of year dummies is included in all specifications. The same notation as in Table 4 is
used. P is the Hansen Sargan test p statistics of over-identifying restrictions, while AR(2) is the Arellano Bond test of secondorder autocorrelation in the errors. sdstockre t is year-within standard deviation of stock return proxying for firm uncertainty, which
is the same with uncertainty t in Table 4, sdallordsre t is standard deviation of the within-year All Ords index returns, which is proxy
for market uncertainty, cov t is the covariance between stock return and market return. avsdwestpac t is within-year standard
deviation of GATCH prediction of Westpac Index measuring macro uncertainty. sdresid t is within-year standard deviation of
residuals of CAPM model of firm stock return, which is the un-predicted part of stock return. All estimates are generated by
Arellano Bond one-step difference GMM. The second to sixth lags of q, cash flow, sales, leverage, year dummies and all the lagged
measures of uncertainty are employed as GMM instruments. *P < 0.1, **P < 0.05; ***P < 0.01.

2014 Economic Society of Australia

100

ECONOMIC RECORD

to idiosyncratic uncertainty rather than aggregate

uncertainty.
VI Conclusion
This paper tests the implication of the Tobins
q model of investment incorporated with uncertainty, using annual data of listed and delisted
non-financial firms on the Australian Stock
Exchange from 1987 to 2009. This is the first
study on the effects of uncertainty on firm
investment in Australia using various measures
of uncertainty. The paper contributes to existing
literature by analysing factors that influence
Australian firms investment.
In the paper, there is strong evidence that
Tobins q, cash flow, sales and leverage are the
driving factors of investment as claimed by
existing literature and as hypothesised in the
empirical model proposed in Section IV. The
implications of the results in this paper are that
financially constrained firms are more sensitive to
uncertainty, while large firms are more likely
influenced by business opportunities (proxied by
q). The contemporaneous and the first lag of the
variables of interest are only significant, deeper
lags are not, implying that firms make investment
decision based on current and the most recent
information.
One of the key implications of this study is
that uncertainty has strong negative effects on
firm investment. The magnitude of the uncertainty effect is relatively strong compared with
other factors of interest; however, the strength
does depend on the measures of uncertainty
utilised. Also, the results show that the impact
of uncertainty is dictated by its type. It appears
that when both firm-specific and market uncertainty are incorporated into the regression
model, the firm-specific uncertainty is more
important for investment decisions compared
with macroeconomic uncertainty; that is, macro
uncertainty does not seem very relevant for firm
investment. This could be due to the weak
proxies used for macro uncertainty or because
the time dummies in the model play an essential
role. Furthermore, there is evidence of a negative association between investment and firm
idiosyncratic uncertainty. This implies that when
the firm becomes more uncertain and as the
volatility of stock price returns increases, companies take more caution and invest less. We
also see a decline in investment when companies
face lower revenues and a higher risk of
borrowing.

JUNE

REFERENCES
Abel, A.B. (1983), Optimal Investment under Uncertainty, American Economic Review, 73, 22833.
Abel, A.B. and Eberly, J.C. (1999), The Effects of
Irreversibility and Uncertainty on Capital Accumulation, Journal of Monetary Economics, 44, 33977.
Arellano, M. and Bond, S. (1991), Some Tests of
Specification for Panel Data Monte-Carlo Evidence
and an Application to Employment Equations,
Review of Economic Studies, 58, 27797.
Barnett, V. and Lewis, T. (1994), Outliers in Statistical
Data. John Wiley, Chichester.
Baum, C.F., Caglayan, M. and Talavera, O. (2008),
Uncertainty Determinants of Firm Investment,
Economics Letters, 98, 2827.
Bernanke, B.S. (1983), Irreversibility, Uncertainty,
and Cyclical Investment, The Quarterly Journal of
Economics, MIT Press, 98, 85106.
Blanchard, O., Rhee, C. and Summers, L. (1993), The
Stock Market, Profit, and Investment, The Quarterly
Journal of Economics, 108, 11536.
Bloom, N., Bond, S. and Van Reenen, J. (2007),
Uncertainty and Investment Dynamics, Review of
Economic Studies, 74, 391415.
Bo, H. (2002), Idiosyncratic Uncertainty and Firm
Investment, Australian Economic Papers, 41, 114.
Bond, S. and Meghir, C. (1994), Dynamic Investment
Models and the Firms Financial Policy, The Review
of Economic Studies, 61, 197222.
Bond, S., Moessner, R., Mumtaz, H. and Syed, M.
(2005), Microeconometric Evidence on Uncertainty and Investment. Institute for Fiscal Studies,
London, UK.
Bontempi, M.E., Golinelli, R. and Parigi, G. (2010),
Why Demand Uncertainty Curbs Investment:
Evidence from a Panel of Italian Manufacturing Firms, Journal of Macroeconomics, 32, 21838.
Bulan, L.T. (2005), Real Options, Irreversible
Investment and Firm Uncertainty: New Evidence
from US Firms, Review of Financial Economics,
14, 25579.
Caballero, R.J. and Pindyck, R.S. (1996), Uncertainty,
Investment, and Industry Evolution, International
Economic Review, 37, 64162.
Carrington, S. and Tran, T.L. (2012), What Determines
Investment in Australia: The Role of the Stock
Market, Uncertainty and Market Imperfection, The
41st Australian Conference of Economists, Melbourne, July 812.
Chang, X., Tan, T.J., Wong, G. and Zhang, H. (2007),
Effects of Financial Constraints on Corporate Policies in Australia, Accounting and Finance, 47, 85
108.
DeMarzo, P. and Sannikov, Y. (2006), Optimal Security Design and Dynamic Capital Structure in a
Continuous-Time Agency Model, Journal of
Finance, 61, 2681724.
Dixit, A. and Pindyck, A. (1994), Investment under
Uncertainty. Princeton University, Princeton, NJ.

2014 Economic Society of Australia

2014

FIRM INVESTMENT AND UNCERTAINTY: AUSTRALIAN EVIDENCE

Episcopos, A. (1995), Evidence on the Relationship

between Uncertainty and Irreversible Investment,
The Quarterly Review of Economics and Finance, 35,
4152.
Fazzari, S.M., Hubbard, R.G. and Petersen, B.C.
(1988), Financing Constraints and Corporate Investment, Brookings Papers on Economic Activity, 1,
141206.
Fuss, C. and Vermeulen, P. (2008), Firms Investment
Decisions in Response to Demand and Price,
European Central Bank Working Paper Series No.
347, European Central Bank, Frankfurt, Germany.
Greenwald, B. and Stiglitz, J.E. (1993), Financial
Market Imperfections and Business Cycles, Quarterly Journal of Economics, 108, 77114.
Guiso, L. and Parigi, G. (1999), Investment and
Demand Uncertainty, Quarterly Journal of Economics, 114, 185227.
Hatakeda, T. (2002), How Do We Account for
the Relationship between Investment and Uncertainty? Nihon University College of Economics,
Tokyo, Japan.
Koetse, M. J. (2006), Determinants of Investment
Behaviour Methods and Applications of Meta-Analysis. Rozenberg Publishers, Amsterdam, the Netherlands.
La Cava, G. (2005), Financial Constraints, the User
Cost of Capital and Corporate Investment in Australia, Reserve Bank of Australia Research Discussion
Paper No. 2005-12, Reserve Bank of Australia,
Canberra, Australia.

2014 Economic Society of Australia

101

Leahy, J.V. and Whited, T.M. (1996), The Effect

of Uncertainty on Investment: Some Stylized
Facts, Journal of Money Credit and Banking, 28,
6483.
Mills, K., Morling, S. and Tease, W. (1995), The
Influence of Financial Factors on Corporate Investment, The Australian Economic Review, 2, 5064.
Panousi, V. and Papanikolaou, D. (2012), Investment,
Idiosyncratic Risk, and Ownership, The Journal of
Finance, 67, 111348.
Pattillo, C. (1998), Investment, Uncertainty and Irreversibility in Ghana, International Monetary Fund
Staff Paper, 45, 522553.
Pindyck, R. (1982), Adjustment Costs, Demand
Uncertainty and the Behaviour of the Firm, American Economic Review, 72, 41527.
Romer, D. (2006), Advanced Macroeconomics, 3rd edn.
McGraw-Hill, New York.
Romer, C. D. and Romer, D. H. (2000), Federal
Reserve Private Information and the Behaviour of
Interest Rates. American Economic Review, 90, 429
457.
Von Kalckreuth, U. (2003), Exploring the Role of
Uncertainty for Corporate Investment Decisions in
Germany. Economic Research Group of the Deutsche
Bundesbank, Frankfurt.
Wooldridge, J. M. (2002), Econometric Analysis of
Cross Section and Panel Data. The MIT Press,
Massachusetts.