23261366

American Finance Association
Real Options, Volatility, and Stock Returns

Author(s): GUSTAVO GRULLON, EVGENY LYANDRES and ALEXEI ZHDANOV
Source: The Journal of Finance, Vol. 67, No. 4 (AUGUST 2012), pp. 1499-1537
Published by: Wiley for the American Finance Association
Stable URL: http://www.jstor.org/stable/23261366
Accessed: 03-04-2017 12:34 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
American Finance Association, Wiley are collaborating with JSTOR to digitize, preserve and extend
access to The Journal of Finance
This content downloaded from 132.174.250.76 on Mon, 03 Apr 2017 12:34:22 UTC
All use subject to http://about.jstor.org/terms
THE JOURNAL OF FINANCE VOL. LXVII, NO. 4 AUGUST 2012
Real Options, Volatility, and Stock Returns
GUSTAVO GRULLON, EVGENY LYANDRES, and ALEXEI ZHDANOV*
ABSTRACT
We provide evidence that the positive relation between firm-level stock returns
firm-level return volatility is due to firms' real options. Consistent with real op
theory, we find that the positive volatility-return relation is much stronger for f
with more real options and that the sensitivity of firm value to changes in volat
declines significantly after firms exercise their real options. We reconcile the evi
at the aggregate and firm levels by showing that the negative relation at the aggr
level may be due to aggregate market conditions that simultaneously affect
market returns and return volatility.
It is well established in the asset pricing literature that aggregat

returns are negatively correlated with aggregate market volatility (e.g
Schwert, and Stambaugh (1987), Campbell and Hentschel (1992), a
(1995)). One possible explanation for this negative relation is the
effect" hypothesis (e.g., Black (1976) and Christie (1982)), which
when stock prices fall, firms become more levered, raising the volatil
returns. Another explanation, due to French, Schwert, and Stambaug
holds that, because an increase in systematic volatility raises risk pre
expected future stock returns, an unexpected change in (systematic) v
likely to reduce firm values, leading to a negative association between
and contemporaneous returns.
*Gustavo Grullon is with the Jesse H. Jones Graduate School of Business, Rice Un
Evgeny Lyandres is with the School of Management, Boston University. Alexei Zhd
the University of Lausanne and Swiss Finance Institute. The authors thank Rui Abuquerque,
Yakov Amihud, Doron Avramov, Clifford Ball, Alexander Barinov, Jonathan Berk, Gennaro Bernile,
Nicolas Bollen, Jacob Boudoukh, Tim Burch, Murray Carlson, Lauren Cohen, Francois Degeorge,
Darrell Duffie, Rafi Eldor, Wayne Ferson, Amit Goyal, Dirk Hackbarth, Campbell Harvey (the
Editor), Ohad Kadan, Markku Kaustia, Matti Keloharju, Timo Korkeamaki, Moshe Levy, Lubomir
Litov, Hong Liu, Roni Michaely, Barb Ostdiek, Dino Palazzo, Brad Paye, Neil Pearson, Gordon
Phillips, Lukasz Pomorski, Amir Rubin, Jacob Sagi, Dan Segal, Anjan Thakor, Yuri Tserlukevich,
Masahiro Watanabe, James Weston, Zvi Wiener, Yuhang Xing, Guofu Zhou, an Associate Editor,
an anonymous referee, and seminar participants at Aalto School of Economics, Hebrew Univer
sity, Interdisciplinary Center Herzliya, Louisiana State University, Rice University, Texas A&M
International University, Vanderbilt University, Washington University at Saint Louis, Univer
sity of Illinois at Urbana-Champaign, University of Miami, University of Texas at San Antonio,
2008 University of British Columbia Winter Finance Conference, 2008 Rotschild Caesarea Center
Conference, 2008 European Finance Association Meetings, 2011 Finance Down Under Conference,
and 2011 Napa Valley Conference for helpful comments and suggestions. The authors also thank
Hernan Ortiz-Molina for help with using union membership data and Sarah Diez for valuable
research assistance.
1499
1500 The Journal of Finance
Contrary to the evidence at the aggregate level, however, Du

that stock returns and volatility are positively correlated
This empirical finding has important theoretical implicati
inconsistent with the leverage and risk premia hypotheses,
the conventional wisdom on the relation between volatility an
The main contribution of our paper is twofold. First, we
nal explanation for the positive contemporaneous relation betw
returns and firm-level volatility documented in Duffee (1995)
vide an explanation for the difference between the aggregate
relation and the firm-level volatility-return relation. Surprisi
importance of these issues to our understanding of the role o
set pricing, research on these topics has been very limited
(2012) and Duffee (2002)).
We hypothesize that the positive relation between firm-level returns and
firm-level volatility may be due to real options that firms possess. One of the
main implications of real options theory is that a real option's value is in
creasing in the volatility of an underlying process (i.e., demand volatility, cost
volatility, or overall volatility of profits). The main rationale for this relation
is that, since firms can change their operating and investment decisions to
mitigate the effects of bad news (e.g., reduce production, shut down operations,
defer investments) and amplify the effects of good news (e.g., expand produc
tion, restart operations, expedite investments), an increase in the volatility of
an underlying process can have a positive effect on firm value. That is, since
operating and investment flexibility increases the convexity of firm value with
respect to the value of its underlying assets, firm value is an increasing func
tion of volatility, due to Jensen's inequality. Therefore, if real options consti
tute a substantial component of firm value, then it is possible that the positive
return-volatility relation documented in Duffee (1995) is driven by the presence
of these options. We test this hypothesis empirically and find numerous pieces
of supportive evidence.
First, we examine whether the value of firms with abundant investment op
portunities is more sensitive to changes in underlying volatility than the value
of assets in place-based firms. The more investment opportunities a firm has,
the more discretion it has with respect to the timing of its investments and
hence the larger the value of its real options. Thus, if the positive relation be
tween returns and changes in volatility is due to the presence of real options,
whose values are increasing in volatility, then the sensitivity of firm value to
volatility should be stronger among firms with more investment opportunities.
Using a battery of proxies for investment opportunities, we find that the pos
itive contemporaneous relation between returns and changes in volatility is
very strong among firms that are likely to have abundant investment oppor
tunities, while it is substantially weaker among assets in place-based firms.
Specifically, we find that the volatility-return relation is stronger among young
firms, small firms, high R&D firms, and high growth firms.
Second, since the value of real options comes from the ability of firm man
agers to change their decisions as new information arrives, we also investigate
Real Options, Volatility, and Stock Returns 1501
whether proxies for operating flexibility can explain cross-sectional differen

in the volatility-return relation. Consistent with the prediction that volatili
creates value when managers have flexibility to alter their investment
operating decisions, we find that the relation between volatility and stock r
turns is stronger among firms with fewer operational constraints (e.g., firm
in nonunionized industries) and firms that appear to be able to respond bett
to resolutions of uncertainty (e.g., firms with higher convexity of value wit
respect to earnings and sales).
Third, we investigate how the sensitivity of firm values to changes in vola
ity evolves as a firm's mix of growth options and assets in place changes ove
time. On the one hand, a firm develops and accumulates real options. On
other hand, it exercises these options by investing when the value of the be
efits from investing is high enough to offset the value of the option to wai
Thus, the sensitivity of firm value to changes in volatility is expected to
increasing as the firm builds up its real options, and it is expected to decline
when the firm exercises (part of) them. To test this prediction, we use spike
in firms' investment levels, issues of seasoned equity, and spikes in exter
financing in general to proxy for instances of real option exercises. Consiste
with the theory, we find that the sensitivity of firm value to changes in und
lying volatility increases prior to real option exercises, drops sharply follow
exercises of real options, and then starts rising again as firms start to b
up new real options. This evidence strongly suggests that part of the positiv
relation between returns and volatility is driven by the effect of volatility
the value of real options.
Fourth, we demonstrate that the volatility-return relation is much strong
in industries that have been shown to have plenty of growth and strate
options (high-tech, pharmaceutical, and biotechnology industries) and h
levels of operating flexibility (natural resources industries). Furthermore, w
perform a within-industry analysis using a sample of oil and gas firms
investigate more deeply the effect of volatility on stock returns. We focus on
and gas firms because they provide a unique setting for testing the predicti
of the real options theory. Since these firms have valuable timing options on
developing their undeveloped proven reserves, one could use their undevelop
and developed reserve estimates as proxies for their mix of real options
assets in place. Using hand-collected data on oil and gas firms' reserves,
find that, consistent with the theory, the return-volatility relation is stron
among firms with a higher proportion of undeveloped reserves.
In addition, we examine the effect of the return-volatility relation on
performance of asset pricing models. Based on the insights of McDonald
Siegel (1985) and Berk, Green, and Naik (1999), Da, Guo, and Jagannath
(2012) argue that, in the presence of real options, the CAPM may expla
the expected returns on a firm's underlying assets but not necessarily t
expected returns on its equity. This is because when firms possess real option
equity risk becomes a nonlinear function of the risk of the underlying asset
Consistent with this argument, Da, Guo, and Jagannathan (2012) show t
the presence of real options seems to explain the poor performance of t
CAPM. We exploit this result to test our main hypothesis. If

important determinant of the positive relation between volat
turns, then the CAPM, or any asset pricing model that does no
options, should perform better for firms with a weak return-
(firms with relatively few real options) than for firms with
volatility relation (firms with abundant real options). Our em
consistent with the real options theory. Using Gibbons, Ro
(1989) test, we find that the CAPM, as well as the Fama and French (1993)
three-factor model, cannot be rejected within a subsample of firms with a rela
tively weak return-volatility relation, but the models are comfortably rejected
within a subsample of firms with a relatively strong return-volatility relation.
In general, we believe that our paper provides an explanation for the positive
contemporaneous relation between firm-level returns and firm-level volatility
documented in Duffee (1995). The sensitivity of the value of real options to
underlying volatility seems to be an important reason for the cross-sectional
variation in the relation between returns and contemporaneous changes in
volatility and for the evolution of this relation around investment and financing
spikes. These findings complement the existing literature examining the effects
of real options on asset prices (e.g., Berk, Green, and Naik (1999) and Carlson,
Fisher, and Giammarino (2004)).
While the real options hypothesis is consistent with the positive relation be
tween volatility and stock returns at the firm level, it cannot explain the nega
tive correlation between these variables at the aggregate level. We argue that
the negative relation between aggregate stock returns and aggregate return
volatility could be driven by an omitted variable problem. Because investors
tend to be more uncertain about future real output growth during economic
downturns (e.g., Veronesi (1999)), periods of high stock return volatility could
coincide with periods of low stock returns even if the direct effect of volatility on
firm value is positive. That is, volatility may increase when stock prices decline
not because the fundamental relation between these variables is negative, but
because both variables are affected by the same underlying macroeconomic fac
tors. Thus, regressing aggregate stock returns on aggregate market volatility
could lead to inferences that are very different from those obtained in a setting
in which aggregate (market) conditions are controlled for.
We address this issue by focusing on firm-level stock returns rather than on
aggregate returns. Using individual stock returns instead of aggregate returns
allows us to control simultaneously for aggregate factors (proxies for market
conditions) and aggregate volatility. Consistent with the aggregate evidence,
regressions of firm-level returns on aggregate volatility alone reveal a strong
negative return-volatility relation. However, once we control for aggregate mar
ket factors (aggregate market returns, HML, and SMB), aggregate volatility
becomes unrelated to stock returns. More interestingly, we find that aggregate
volatility has a positive effect on the value of real options-based firms and a
negative effect on the value of assets in place-based firms after controlling for
market conditions. These results seem to reconcile the aggregate-level nega
tive relation between volatility and returns with the positive relation at the
firm level.
The remainder of the paper is organized as follows. In Section I we dis

cuss and motivate our measure of volatility, summarize the data, and presen
selected summary statistics. In Section II we estimate the relation between
returns and changes in volatility for subsets of firms characterized by different
mixes of real options and assets in place. Section III investigates the effect
of operating flexibility on the volatility-return relation. Section IV examines
the evolution of the relation between returns and changes in volatility around
times of significant changes in firms' mix of real options and assets in place.
In Section V we perform an industry-level analysis of the relation between
returns and contemporaneous changes in volatility. In Section VI we examine
the effect of the volatility-return relation on the performance of asset pricing
models. Section VII provides evidence that aggregate volatility is unrelated to
stock returns after controlling for underlying market conditions and examines
the relation between returns and aggregate volatility for subsamples of real
options-based and assets in place-based firms. In Section VIII we discuss the
results of robustness checks. Section IX summarizes and concludes.
I. Measure of Volatility, Data Sources, and Summary Statistics
A. Measure of Volatility
Theoretically, the value of a firm's real options is increasing in the volatility
of an underlying process (e.g., McDonald and Siegel (1986)). However, many
aspects of uncertainty regarding potential projects, which include but are not
limited to demand shocks (changes in consumer tastes), supply shocks (changes
in production technologies), and institutional changes, are unobservable. More
over, even if the realizations of these shocks were observable ex post, their
expectations, which affect the value of real options, would not be known. If
stock prices incorporate the value of real options, then the volatility of stock
prices is expected to be related to the volatility of the underlying valuation pro
cesses. This justifies the use of measures of stock return volatility as proxies
for underlying volatility, as in Leahy and Whited (1996) and Bulan (2005).
We follow Ang et al. (2006, 2009) and Duffee (1995), among others, and
estimate firm Vs volatility during month t as the standard deviation of the
firm's daily returns during month t,
E(n,r-/v)2
VOLi t = , (i)
' Nj rit - 1
where rl%x is the natural logarithm of day ret gross excess return on firm
i's stock, is the mean of the logarithms of gross daily returns on firm i's
stock during month t, and rit is the number of nonmissing return observations
during month t. We use logarithmic returns to mitigate the potential mechan
ical effect of return skewness (see Duffee (1995) and Kapadia (2007)) on the
relation between returns and contemporaneous return volatilities. The change
in volatility in month t, AVOLlt, is computed as the difference between the
estimated volatility in month t and the estimated volatility in
AVOLu = VOLlt - YOU,,_x. (2)
B. Main Data Sources and Summary Statistics

We obtain daily stock returns, used in estimating vo
loadings, and monthly returns, used as the dependen
gressions, from CRSP daily and monthly return files,
monthly factor returns and risk-free rates are from
(http:// mba.tuck.dartmouth.edu/ pages/faculty/ken.fre
The time frame of our analysis is from January 1964 to December 2008.
Following Ang et al. (2006), among others, we eliminate utilities (SIC codes
between 4900 and 4999) and financials (SIC codes between 6000 and 6999)
from the sample. Our sample contains over 3 million monthly observations
with nonmissing returns and volatility estimates.
Accounting variables used to compute firm characteristics, measures of in
vestment opportunities (firm size, R&D expenditures, and sales growth), mea
sures of operating flexibility (sensitivity of firm value to profits and sales), as
well as measures of investment and financing spikes, are from COMPUSTAT.
We obtain firms' founding and incorporation years, used to compute firm
age, from Boyan Jovanovic's website (www.njoa.edvi/econ/user/jovanovi). Dates
of firms' earnings announcements used to estimate some of the measures
of operating flexibility are from I/B/E/S. Data on membership in labor
unions are obtained from the Union Membership and Coverage database
(www.unionstats.com), described by Hirsch and Macpherson (2003).1 We ob
tain data on seasoned equity offerings (SEOs), used in our event-time tests,
from Thomson Financial's Securities Data Company. Finally, for our analysis
of the effects of real options on the return-volatility relation in the oil and
gas industry, we hand-collect data on developed and total proven oil and gas
reserves for 72 oil firms between 1995 and 2009 from firms' annual reports.
We present summary statistics for returns, return volatilities, and changes in
return volatility in Table I, which also includes summary statistics for measures
of investment opportunities and managerial flexibility, which we discuss below.
The mean excess return in our sample is 0.6% per month or about 7.2% per year.
The mean (median) daily firm-level stock return standard deviation is 3.17%
(2.45%). Our firm-level volatility estimates using daily data are similar to those
reported in Ang et al. (2006). The small positive mean change in volatility
(0.007%) is consistent with the positive time trend in volatility (e.g., Campbell
et al. (2001) and Cao, Simin, and Zhao (2008)). The standard deviation of the
month-to-month change in return volatility is 1.83%.
1 These data use the Census Industry Classification (CIC) to classify industries, which we map
into Standard Industry Classification (SIC) codes and North American Industry Classification
System (NAICS) in order to match them with Compustat and CRSP.
Table I
Summary Statistics
This table presents summary statistics for returns, measures of volatility and changes in these
measures, as well as for measures of investment opportunities and operating flexibility used
in cross-sectional tests. Returns data are from CRSP. Accounting data are from COMPUSTAT.
The sample period is 7/1963 to 12/2008. A stock's excess return is the difference between its
monthly return and monthly risk-free return. Volatility and its change refers to monthly volatility
of logarithmic daily returns. Book assets are in millions of dollars. Age is the difference between the
current year and the founding year, incorporation year, or the first year that the firm's stock appears
in monthly CRSP files, in that order of availability. R&D/assets is the ratio of R&D expenditures
and lagged book assets. Future sales growth is defined as the difference between sales 4 years
after the year of the observation over sales in the year following the year of the observation divided
by sales in the year following the year of the observation. Earnings convexity is the estimated
coefficient on the squared earnings surprise in the firm-level time-series regression of returns on
earnings announcement days. Sales convexity is the estimated coefficient on the squared sales
surprise in the firm-level time-series regression of returns on earnings announcement days. Union
membership is at the industry level and is obtained from the Union Membership and Coverage
database.
Mean SD P5 Median P95 # Obs.
Excess return 0.603 17.731 -21.709 -0.334 27.442 3,046,524

Volatility 3.172 2.539 0.669 2.453 8.443 3,070,953
A Volatility 0.007 1.831 -2.884 -0.011 3.159 3,040,952
Book assets 2,983.05 31,470.78 3.91 107.65 8,867.75 2,461,941
Age 28.383 31.915 1.000 13.322 96.864 3,062,722
R&D/assets 0.107 0.292 0.000 0.029 0.446 1,067,418
Future 3-year sales growth2.818 19.982 -0.993 0.264 5.888 1,805,829
Earnings convexity 0.0268 1.2252 -1.2928 0.0044 1.3480 849,827
Sales convexity 0.2325 5.3051 -3.1685 0.0000 3.3917 746,283
Union membership 0.106 0.117 0.009 0.060 0.364 1,315,294
II. Return-Volatility Relation and Investment Opportunities

We begin by verifying that the positive relation between firm-level volatility
and firm-level returns documented in Duffee (1995, 2002) and Albuquerque
(2012) holds in our sample. In particular, we estimate monthly cross-sectional
Fama and MacBeth (1973) regressions of individual firm returns, rl t, net of
the risk-free rate, rf<t, on contemporaneous changes in firm-level volatility,
AVOLi t, and a vector of firm characteristics, xl t, most of which are known at
the beginning of month t:
1"i,t ~ rf,t = <*t + Pt&VOLi t + YtVMKTi.t + &t %i,t + i,t- (3)
Following common practice in the asset pricing literature (e.g., Fama and
French (1993), Jegadeesh and Titman (1993), and Cooper, Gulen, and Schill
(2008) among many others), these characteristics are log market equity, log
book-to-market, and past returns. Following Fama and French (1993), we
measure the market value of equity as the share price at the end of June
times the number of shares outstanding. Book equity is stockholders' equity
minus preferred stock plus balance sheet deferred taxes and investment tax
Table II
Returns and Contemporaneous Changes in Volatility

This table presents regressions of firm-level excess returns on the estimated loading on the market
factor, beginning-of-year log book-to-market ratio, log market equity, 6-month lagged return for
months -7 to -2 relative to the observation month, monthly trading volume normalized by the
number of shares outstanding, and month-to-month change in firm-level volatility, AVOLij. The
sample period is 7/1963 to 12/2008. We estimate the regressions monthly and report time-series
means of coefficient estimates along with f-statistics obtained using Newey-West autocorrelation
and heteroskedasticity-consistent standard errors of monthly coefficient estimates in parentheses.
R2 refers to the average monthly R2.
Market factor loading 0.298 0.217 0.300 0.2155 0.171 0.1075

(5.67) (5.03) (5.91) (5.25) (4.04) (3.06)
Log(B/M) 0.413 0.383 0.406 0.380 0.770 0.717
(3.85) (3.68) (3.87) (3.74) (9.32) (8.79)
Log(equity value) 0.141 0.130 -0.152 -0.138 -0.142 -0.1354
(-2.71) (-2.59) (-2.93) (-2.78) (-2.91) (-2.89)
Lag(6-month return) 0.0055 0.0057 -0.0015 -0.0006
(4.41) (4.37) (-1.15) (-0.52)
Volume 2.972 2.476
(10.92) (11.30)
A Volatility 1.308 1.315 1.186
(10.98) (11.01) (10.71)
R2 0.036 0.071 0.041 0.077 0.076 0.105
# Months 539 539 539 539 539 539
credit if available, minus post-retirement benefit assets if available. If stock

holders' equity is missing, we use common equity plus preferred stock par
value. If these variables are missing, we use book assets less liabilities. Pre
ferred stock is preferred stock liquidating value, preferred stock redemption
value, or preferred stock par value in that order of availability. To compute the
book-to-market ratio, we use the December closing stock price times number
of shares outstanding. We match returns from January to June of year t with
COMPUSTAT-based variables of year t 2 , while the returns from July until
December are matched with COMPUSTAT variables of year t 1. Past returns
are defined as buy-and-hold returns for six months over months [t 7, t 2],
In addition, following Karpoff (1987), we include contemporaneous trading
volume, normalized by the number of shares outstanding. The estimated coef
ficient on the market portfolio return, rfMKTit in equation (3), is obtained from
the following regression:
t 7*f x ~t~ VMKTi t^Tm,r ^f.r) 4" (4)

where r; T is the return of firm i in day r belonging to month t, rf T is the
daily risk-free rate, and rm>1 is the daily return on the value-weighted mark
portfolio.
The results of estimating equation (3) are presented in Table II.2
2 Here and below, the standard errors of the Fama and MacBeth (1973) estimates are c
puted using Newey and West (1987) heteroskedasticity- and autocorrelation-consistent varia
covariance matrices.
This table shows that contemporaneous changes in firm-level volatility are p

itively related to stock returns, the relation being highly statistically signific
in all specifications. As expected, the coefficients on the market factor load
and on log book-to-market are significantly positive, while the coefficients on
size are significantly negative in all specifications. The coefficients on conte
poraneous volume are positive and highly statistically significant, consist
with Karpoff (1987). Interestingly, while the coefficients on past returns ar
positive and significant in regressions in which trading volume is exclud
they become insignificant with the inclusion of contemporaneous trading vo
ume. This result is consistent with Cooper, Gulen, and Schill (2008), who fin
that the coefficient on past returns is sensitive to the set of other independ
variables included in return regressions.
Perhaps the most common type of real option is the option to invest (e
Brennan and Schwartz (1985), McDonald and Siegel (1986), Majd and Pind
(1987), and Pindyck (1988), among many others). Therefore, one way to exam
ine whether the relation between firms' stock returns and contemporane
changes in return volatility is driven by the effect of volatility on the valu
real options is to compare the strength of this relation across subsample
firms with different mixes of investment opportunities and assets in place.
To analyze the effects of investment opportunities on the return-volatility
relation, at the end of each year we sort firms based on measures of investme
opportunities and form investment opportunity-based quintiles. We use four
measures of investment opportunities. The first (inverse) measure is firm si
Larger firms tend to have larger proportions of their values represented
assets in place, while smaller firms tend to rely more heavily on investm
opportunities (e.g., Brown and Kapadia (2007)). We define firm size as the boo
value of firm assets.
Our second (inverse) proxy for investment opportunities is firm age. Older,
more established firms tend to have larger proportions of their value repre
sented by existing assets (Lemmon and Zender (2010)). As in previous studies,
we define a firm's age as the difference between current year and founding
year, incorporation year, or the first year in which the firm's stock appears in
monthly CRSP files, in that order of availability.
The third investment opportunity proxy that we use is R&D intensity. Since
research and development generates investment opportunities, the larger the
firm's relative R&D expenditures, the more real options the firm is expected to
have. R&D intensity is defined as the ratio of annual R&D expenditures and
beginning-of-year book assets.
Our fourth measure of investment opportunities is future sales growth. An
increase in sales (and production) in the future is likely to be caused by the
future exercise of real options. The clear drawback of future sales growth as a
measure of current investment opportunities is that it suffers from the look
ahead bias. However, it can still be useful in our setting because the regressions
we estimate are not predictive regressions. Instead, our tests focus on the con
temporaneous relation among the variables of interest. Thus, we use realized
future sales growth as an instrument for expected sales growth. In order not
to induce spurious correlation caused by contemporaneous surprises to sales
and to firm value, we measure future sales growth starting fr

the period for which the relation between returns and change
estimated. Specifically, future sales growth is defined as the di
sales 4 years after the year of the observation over sales in th
the year of the observation divided by sales in the year follow
the observation.
In Table III we estimate the following Fama-MacBeth cross-sectional
regressions:
ri,t rf,t = ott + faAVOLij + vtGRijAVOLi t + YtVMKTi,, + &t%i,t + (5)
where GRltAVOLlt is the product of AVOLit and one of the four investment
opportunity measures: log size, log age, log R & D to assets ratio, and future
sales growth. To allow for an intuitive interpretation of the results, we nor
malize each of the investment opportunity measures by subtracting its sample
mean and dividing the resulting difference by its in-sample standard deviation.
The rest of the variables are as in equation (3).
In the first column, in which size is used as a proxy for the relative amount
of investment opportunities, the mean estimate of ft, which is interpreted
as the sensitivity of firm value to changes in volatility for a firm whose size
is equal to the sample mean, is 0.97 and is highly statistically significant.
The coefficient on the interaction between normalized log size and A VOL
equals 0.52, implying that a one standard deviation reduction in log size
from the sample mean is associated with a 0.52 increase in the return-A VOL
relation. Returns of firms whose book assets are two standard deviations above
the sample mean are not related to changes in volatility, while returns of firms
whose assets are two standard deviations below the mean are twice as sensitive
to changes in volatility as those of firms with mean size. To put it differently,
a two-standard-deviation positive shock to volatility would result, on average,
in a 7.3% monthly return for small firms, while it would have no effect on the
value of large firms. The coefficient on the interaction between A VOL and log
size is not only economically large, but also highly statistically significant.
The results reported in the second column, in which we use log age as an
inverse proxy for the availability of investment opportunities, are also consis
tent with real options theory. The coefficient on the interaction between A VOL
and log age is 0.12 and is much smaller in magnitude than the corresponding
coefficient in the first column. However, this coefficient is still economically
large: the relation between A VOL and returns is 48% stronger on average for
a firm whose age is two standard deviations below the sample mean than for a
firm whose age is two standard deviations above the sample mean. The results
using log R&D intensity, reported in the third column, are quite similar. A one
standard deviation increase in R & D expenditures from the sample mean is
associated with an approximately 11% increase in the sensitivity of firm value
to volatility. Finally, the evidence based on future sales growth, reported in
the fourth column, indicates that the relation between returns and changes in
volatility is significantly stronger for firms with high future sales growth than
Table III
Returns, Contemporaneous Changes in Volatility, and Investment

Opportunities
This table presents regressions of firm-level excess returns on the estimated loading on the mar
ket factor, beginning-of-year log book-to-market ratio, log market equity, 6-month lagged return
for months -7 to -2 relative to the observation month, monthly trading volume normalized by
the number of shares outstanding, month-to-month change in firm-level volatility, AVOLlt, and
an interaction variable equal to the product of AVOLiand one of the four investment oppor
tunity measures: log (book assets), log (age), log (R&D to assets), and future sales growth. See
Section III for definitions of the investment opportunities measures. We normalize each of the
investment opportunity measures by subtracting its in-sample mean and dividing the difference
by in-sample standard deviation. The sample period is 7/1963 to 12/2008. We estimate the regres
sions monthly and report time-series means of coefficient estimates along with f-statistics obtained
using Newey-West autocorrelation- and heteroskedasticity-consistent standard errors of monthly
coefficient estimates in parentheses. R2 refers to the average monthly J?2. We also present a sum
mary of the economic effects of investment opportunity measures on the AVOLj.f-return relation.
Low (high) investment opportunities means that the effect of AVOLi t on returns is calculated
using the value of the proxy for investment opportunities at two standard deviations below (above)
the mean. Large AVOLi t refers to AVOLi t that is two standard deviations above the mean; such
change equals 3.66%.
Proxy for Investment Opportunities
Future Sales
Book Assets Age R&D/Assets Growth
Market 0.120 0.106 0.147 0.134

(3.43) (3.01) (3.86) (3.83)
Log(B/M) 0.733 0.709 0.703 0.637
(9.05) (8.72) (7.51) (7.82)
Log(equity value) -0.128 -0.139 -0.155 -0.174
(-2.72) (-2.99) (-3.28) (-3.73)
Lag( 6-month return) -0.001 -0.001 -0.001 -0.001
(-0.61) (-0.47) (-0.39) (-0.44)
Volume 2.762 2.752 2.517 2.488
(11.27) (11.19) (1.0.35) (10.76)
A Volatility 0.967 1.213 1.102 1.220
(9.02) (10.71) (10.39) (10.41)
A Volatility * investment opportunities -0.515 -0.118 0.120 0.273
(-11.08) (-3.64) (1.77) (3.52)
R2 0.109 0.107 0.123 0.108
# Months 539 539 539 521
Total effect of AVOLl t on return
High investment opportunities 1.997 1.449 1.342 1.766
Low investment opportunities -0.063 0.977 0.862 0.674
Impact of large A VOLi t on return
High investment opportunities 7.31% 1.73% 4.91% 6.47%
Low investment opportunities -0.23% 3.58% 3.16% 2.47%
it is for firms with lower growth. Returns of firms with future sales growth
two standard deviations above the sample mean are 160% more sensitive to
changes in volatility on average than returns of firms whose future growth in
sales is two standard deviations below the sample mean.
Overall, the results in Table III demonstrate that the value of firms with
characteristics that are likely to be related to more abundant and valuable
investment opportunities are more sensitive to changes in return volatility. If
return volatility is correlated with the volatility of processes underlying firms'
real options, as argued in Leahy and Whited (1996) and Bulan (2005), then this
evidence is consistent with the real options-based explanation for the positive
relation between returns and return volatility at the firm level.
III. Return-Volatility Relation, Convexity, and Flexibility

Real options take different forms and are not limited to investment opportu
nities. In this section we go beyond examining the relation between investment
opportunities and the return-volatility relation. We construct additional prox
ies for the relative importance of real options for firm value and examine the
return-A VOL relation across different real options subsamples.
We begin with the simple observation that if a firm has real options, then
its value function is convex in the process underlying the options. The reason
is that, in real options models, managerial flexibility is what generates the
convexity of the value function with respect to the value of the underlying
process (e.g., Brennan and Schwartz (1985), McDonald and Siegel (1986), Majd
and Pindyck (1987), and Pindyck(1988)).3 Therefore, due to Jensen's inequality,
the sensitivity of firm value to the volatility of its underlying assets should
be increasing in the firm's flexibility to alter its operational and investment
decisions (i.e., increasing in the convexity of the firm's value function (3 ^j2x>)),
1 d2V(x)
E[V(*)]=E [VCe)] + -^
2 dxz
(6)
where V (x) is firm value as a function of variable x having a standard deviation

of ox.
While we do not know what this underlying process is for any given firm,
the theoretical real options literature provides some guidance. In many real
options models the underlying process is either firm earnings (e.g., McDonald
and Siegel (1986) and Pindyck (1993)) or demand for a firm's product (e.g.,
Caballero and Pindyck (1992)). For example, it is well established that the
source of convexity of the profit function in prices comes from the ability of
firms to adjust output according to market conditions (e.g., Marschak and
Nelson (1962) and Oi (1961)). Without this flexibility, profits would be linear
in prices. Mills (1984) formalizes this argument and shows that the expected
profit of a competitive firm in a market in which prices are stochastic equals
E(7r(p)) = 7T(E(p)) + (7)
3 The presence of leverage can also generate convexity in the value function (e.g., Galai and
Masulis (1976)). However, we show below that our results are not driven by the leverage effect.
where nip) denotes profit as a function of price p having variance er|, and y is a
parameter of the profit function that is inversely related to the cost of altering
output in response to price changes. Based on the intuition of Stigler (1939), y
is typically referred to as a measure of flexibility. Equations (6) and (7) show
that the strength of the effect of volatility on a firm's expected profit and firm
value is increasing in its flexibility.
We use two approaches to measuring flexibility. First, following the intuition
in Bernardo and Chowdhry (2002), we use the convexity of a firm's value in
its earnings and in its sales as proxies for operating flexibility. The economic
motivation is that if a firm is able to expand operations during good times and
contract operations during bad times, then its value would be a convex function
of its underlying economic process (e.g., profits, sales). In the second approach
we use the level of union membership in the firm's industry as an inverse proxy
for operating flexibility. The idea is that, since the existence of unions hinders
firms' ability to adjust their workforce in response to changes in economic
conditions (e.g., Abraham and Medoff (1984), Gramm and Schnell (2001), and
Chen, Kasperczyk, and Ortiz-Molina (2011)), one would expect firms in highly
unionized industries to be less flexible.
To estimate the convexity of firm value in its earnings, we focus on earnings
announcement days, obtained from I/B/E/S, and for each firm-announcement
day observation that occurs in quarter t we estimate the following firm-level
time-series regression using data in quarters r e(t - 20, t - 1):
Abnj-i,r = oiij + PijEarrusurpi x + Yi,tEarnJiurpi x2 + eia, (8)

where Abn-TiiX is the return on firm i's stock on the earnings announcement
day in quarter r net of its expected return on that day, which equals its beta
estimated using equation (4) in the month preceding the month of the earnings
announcement times the return on the market portfolio on the earnings an
nouncement day, and Earn^surpi T is the "standardized unexpected earnings"
(SUE).
To estimate SUE, we follow the procedure in Bernard and Thomas (1989)
and Brandt et al. (2009). Specifically, we define earnings surprises as
Earnit% E(Earriix)
Earnsurpi T = , (9)
' o(Earnix)
where Earnia is the firm's earnings per share in qua

x expected earnings per share, and a{Earriix) is t
Earrn x over quarters (t 8, r 1). Expected earn
using a seasonal random walk model:
8
ECEarnitX) = Earni:X-i + y^(Earni<r-n - Earnitr-n-4)/8. (10)

n 1
The measure of convexity of firm value to its earnings surprise is the es

timated coefficient on the squared earnings surprise, y^t in equation (8). The
more convex firm i's value in its earnings is, the larger y^t is,
more strongly to good news than to bad news, and the larger th
value that is likely to be attributable to real options. As in th
we estimate cross-sectional Fama-MacBeth regressions simil
in which the interaction variables are constructed as the prod
normalized convexity estimates.
Column 1 of Table IV reveals that the relation between returns and contem
poraneous changes in volatility is increasing in the estimated convexity of firm
value in its earnings. The coefficient on the interaction between A VOL and nor
malized earnings convexity is equal to 0.052. This implies that a one standard
deviation increase in earnings convexity from the sample mean is associated
with an 8% increase in the magnitude of the return-A VOL relation. In other
words, the sensitivity of firm value to volatility for a firm whose earnings con
vexity is two standard deviations above the sample mean is 38% higher than
the return-AVOL relation for a firm whose earnings convexity is two standard
deviations below the sample mean.
To estimate the convexity of firm value in demand shocks, we follow the
empirical industrial organization literature (e.g., Ghosal (1991) and Guiso and
Parigi (1999)) and use sales as a proxy for demand. The regression we estimate
is similar to equation (8),
Abnj-it = ctij + Pi,tSalesjsurpix + Yi,tSalesJsurpix2 + e;,T, (11)
where Salesjsurpi r is estimated similar to Earnjsurpl T in equation (9). The

regressions involving sales convexity are reported in the second column of
Table IV. The results are somewhat stronger than those based on earnings con
vexity. The coefficient on the interaction between A VOL and sales convexity is
larger than the corresponding coefficient on the interaction between A VOL and
earnings convexity, and it is statistically significant. The return-A VOL relation
is 51% stronger for firms whose sales convexity measure is two standard devi
ations above the sample mean than for firms whose sales convexity measure is
two standard deviations below the sample mean.
As discussed above, we also use labor union membership in a firm's industry
as an inverse proxy for its operating flexibility. Chen, Kasperczyk, and Ortiz
Molina (2011) argue that firms with a highly unionized workforce face obstacles
when trying to reduce the workforce in bad economic times. We therefore expect
the return-AVOL relation to be inversely related to the level of unionization.
The results are presented in the third column of Table IV. The mean coefficient
on the interaction between A VOL and the normalized union membership rate
equals 0.12. This coefficient is not only statistically significant but is also
economically large as it implies that the sensitivity of value to volatility for a
firm in an industry with union membership two standard deviations above the
sample mean is twice as large as the sensitivity for a firm in an industry with
union membership two standard deviations below the sample mean. This result
is consistent with the hypothesis that firms operating in industries with higher
union membership rates have lower flexibility, resulting in a lower proportion
Table IV
Returns, Contemporaneous Changes in Volatility, and Measures

of Flexibility
This table presents regressions of firm-level excess returns on the estimated loading on the market
factor, beginning-of-year log book-to-market ratio, log market equity, 6-month lagged return for
number of shares outstanding, month-to-month change in firm-level volatility, AVOLij, and an
interaction variable equal to the product of AVOLi t and one of the three flexibility measures:
earnings convexity, sales convexity, and union membership. See Section IV for details on esti
mation of the flexibility measures. We normalize each of the flexibility measures by subtracting
its in-sample mean and dividing the difference by the in-sample standard deviation. The sample
period is 7/1985 to 12/2008. We estimate the regressions monthly and report time-series means
of coefficient estimates along with ^statistics obtained using Newey-West autocorrelation- and
heteroskedasticity-consistent standard errors of monthly coefficient estimates in parentheses. R2
refers to the average monthly R2. We also present a summary of the economic effects of the flexibil
ity measures on the A VOLj.creturn relation. Low (high) flexibility means that the effect of AVOLj t
on returns is calculated using the value of the proxy for flexibility at two standard deviations below
(above) the mean. Large A VOLi t refers to A VOLi t that is two standard deviations above the mean;
such change equals 3.66%.
Proxy for Flexibility
Earnings Convexity Sales Convexity Union Membership
Market 0.103 0.112 0.161

(1.99) (2.14) (2.31)
Log(B/M) 0.557 0.525 0.750
(4.85) (4.78) (2.26)
Log(equity value) -0.196 -0.188 -0.165
(-3.22) (-3.14) (-2.51)
Lag(6-month return) -0.000 -0.000 -0.002
(-0.14) (-0.13) T <1 o
Volume 1.251 1.211 1.218
(8.21) (8.29) (7.30)
A Volatility 0.654 0.649 0.692
(6.92) (7.00) (6.87)
A Volatility * flexibility 0.052 0.066 -0.121
(1.92) (2.36) (-3.84)
R2 0.077 0.079 0.086
# Months 287 287 287
Total effect of AVOLrt on return
High flexibility 0.758 0.781 0.934
Low flexibility 0.550 0.517 0.450
Impact of large AVOL,t on return
High flexibility 2.78% 2.86% 3.42%
Low flexibility 2.01% 1.89% 1.65%
of their value represented by real options and in a lower sensitivity of their

value to changes in volatility.
To summarize, the results in this section are consistent with the real
options-based explanation for the positive return-AVOL relation at the firm
level. Firms whose value is more convex in their earnings and sales and firms
operating in industries with lower union membership are more sensitive to

changes in volatility.
IV. Evolution of the Return-Volatility Relation around Real Option

Exercises
The tests in the previous two sections rely on comparing the sensitiv
firm value to changes in return volatility across different sets of firms
measures of real options. In this section we perform an alternative tes
hypothesis that the positive relation between firm-level returns an
poraneous changes in volatility is driven by real options. This test is b
the time-series evolution of the return-A VOL relation for firms expe
shocks to their mix of real options and assets in place, which occur ar
options exercises.
Firms exercise many of their real options by investing. A spike in a
real investment rate can signal an exercise of investment options.
examine changes in the return-A VOL relation around investment
investment opportunities constitute a significant component of firm va
this relation is expected to be decreasing following their exercise.
We use years of abnormally high investment activity as a proxy for
ment spikes. Following Whited (2006), we define a firm-level investme
as a year in which the firm's investment rate exceeds three times its
annual investment rate throughout the sample period. There are 18,65
ment spike years in our sample. For each firm-year we compute th
timing of the previous spike and the next spike and form five su
consisting of firm-years with 2 years prior to the next spike, 1 year
the spike, the spike year, 1 year after the spike, and 2 years post inve
spike.4
Panel A of Table V reports the coefficients on A VOL, from regressions as in
equation (5), estimated for 5 years around investment spikes. The numbers
in brackets in the fourth column denote ^-statistics from the Wald test of the
equality of the A VOL coefficients in years -1 and 1 relative to an investment
spike.The mean coefficient on A VOL stays roughly constant prior to the invest
ment spike year (from year -2 to year -1). The coefficient on A VOL decreases
around the investment spike (from year -1 to the spike year, and to year 1),
and then rises again (from year 1 to year 2 relative to the investment spike).
The differences in the return-A VOL relation between years -1 and 1 are statis
tically significant and economically meaningful. The sensitivity of firm value
to changes in volatility is almost twice as high in the year preceding an invest
ment spike, during which firms are likely to exercise part of their real options,
than in the year following the investment spike year.
4 We exclude firm-years for which we cannot unambiguously assign the timing relative to a
spike. This occurs for firm-years with fewer than 3 years since the last spike and fewer than
3 years until the next spike.
Table V
The Relation between Returns and Changes in Volatility around

Investment Spikes
This table presents regressions of firm-level excess returns on the market factor, beginning-of
year log book-to-market ratio, log market equity, 6-month lagged return for months -7 to -2
relative to the observation month, monthly trading volume normalized by the number of shares
outstanding, and month-to-month change in firm-level volatility, AVOLi t, estimated separately
for subsamples of firms-months belonging to years [-2, 2] relative to an investment-spike year,
SEO, or a financing spike year. Specifically, for each firm-month we compute the relative timing
of the previous spike (SEO) and the next spike (SEO) and form five subsamples (four subsamples
for SEO firms), consisting of firms with 2 years to the next spike, 1 year to the next spike, the
spike year, 1 year after the spike, and 2 years after the spike (2 years before the SEO and 2 years
after). Panel A presents results for investment spike firms. An investment spike is a firm-year in
which a firm's investment rate is at least three times higher than its time-series median. Panel
B presents results for SEO firms. The date of an SEO is its filing date, or issue date if filing date
is unavailable. Panel C presents results for financing spike firms. Financing spikes are firm-years
in which the combination of net new issues of equity and debt exceed 10% of beginning-of-year
book assets. The sample period is 1/1970 to 12/2008. We estimate the regressions monthly and
report time-series means of the coefficient on the change in volatility estimates along with their
i-statistics obtained using standard errors of monthly coefficient estimates in parentheses. The
numbers in brackets are the i-statistics for the differences between the estimated coefficient on
the change in volatility in year 1 and that in year -1 relative to the investment spike, SEO, or
financing spike. We report similar results for matched firms, defined in Section V. The difference
row refers to the mean change in the coefficient on change in volatility between year -1 and year
1 for sample firms minus that for matched firms, with ^-statistics in parentheses.
Panel A: Investment spikes
Years since
investment spike -2 -1 0 1 2
Issuing firm 1.232 1.254 1.021 0.691 1.100

(3.78) (7.97) (7.70) (5.78) (6.56)
[2.85]
Matched firm 0.769 0.867 0.860 0.803 0.812
(3.12) (6.31) (6.76) (5.69) (5.32)
[0.32]
Difference 0.499
(1.78)
Panel B: SEOs
Years since SEO
Issuing firm 1.035 1.874 0.212 0.572

(3.28) (13.65) (1.40) (2.71)
[8.13]
Matched firm 1.227 1.320 1.173 1.235
(3.65) (10.14) (9.01) (9.55)
[0.80]
Difference 1.515
(5.51)
(continued)
Table VContinued
Panel C: Financing spikes

Years since
financing spike -2 -1 0 1 2
Issuing firm 0.724 1.253 0.711 0.583 1.270

(6.22) (8.67) (6.65) (4.98) (7.22)
[3.60]
Matched firm 0.783 0.822 0.798 0.766 0.771
(5.76) (6.43) (6.87) (5.73) (4.75)
[0.30]
Difference 0.614
(2.33)
However, this drop could also be consistent with the following alternative
explanation. There is typically a stock price run-up prior to exercises of in
vestment opportunities. In addition, large investments are typically followed
by low returns (e.g., Anderson and Garcia-Feijoo (2006), Lyandres, Sun, and
Zhang (2008), and Xing (2008)). In our sample, the mean returns of firms expe
riencing an investment spike are 15% and 20% in years -2 and -1 relative to
the spike, compared with the mean return of 10% in the year of the spike and
mean returns of-4% and 3% in years 1 and 2 relative to the spike. Despite the
fact that we use logarithmic returns while estimating return volatilities and
changes in return volatility, high returns may be associated with high return
volatilities, causing a mechanical drop in the sensitivity of returns to changes
in volatility that coincides with the drop in average stock returns following in
vestment spikes. To ensure that this drop in the return-AVOL relation around
investment spikes is not fully explained by the potentially spurious relation
between returns and return volatilities, we present evidence on the evolution
of the return-A VOL relation for the sample of matched firms that did not ex
perience spikes in their investment activity.
Specifically, for each of the 5 years around an investment spike, we find a
matched firm that satisfies the following criteria. First, the matched firm has
to belong to the same quintile of beginning-of-year book assets and book-to
market as the firm experiencing an investment spike.5 Second, to ensure that
the matched firm does not have abnormal investment intensity, we require it
to have an investment rate below its time-series median. Out of the set of firms
satisfying these two criteria, we choose the firm with the annual stock return
closest to that of the investment spike firm. As a result, for each firm experienc
ing an investment spike, we find firms with relatively low investment rates,
similar characteristics, and similar returns in each of the years around the
investment spike. This procedure allows us to separate the possible mechan
ical relation between returns and changes in return volatility caused by the
5 We match firms based on their size and book-to-market ratio following Barber and Lyon (1997)
and Lyon, Barber, and Tsai (1999).
evolution of returns around investment spikes from the effect of the exercise of
real options on the sensitivity of firm values to changes in volatility.
The evolution of the coefficient on A VOL within the matched firm sample is
also presented in Table V. The results indicate that a relatively small portion of
the reduction in the sensitivity of firm values to volatility is attributable to the
mechanical effect discussed above. The coefficient on A VOL within the sample
of comparable firms decreases by less than 10% from year -1 relative to the
investment spike to year 1 relative to the spike, compared with a reduction of
close to 50% for firms going through investment spikes, the difference being
statistically significant.
One potential drawback of examining the behavior of the return- A VOL re
lation around investment spikes is the continuous nature of investments. The
exercise of an investment opportunity might not coincide with an investment
spike. Thus, we attempt to estimate the timing of the decision to exercise real
options and to transform them into assets in place. Firms frequently raise ex
ternal funds in order to finance their investments. Real options models (e.g.,
Carlson, Fisher, and Giammarino (2006, 2010)) regard seasoned equity offer
ings (SEOs) as a signal of the decision to exercise growth options by investing
the SEO proceeds. Consistent with these models, Lyandres, Sun, and Zhang
(2008) find that investment rates of firms that issue equity are significantly
higher than those of similar non-issuers. Thus, we supplement the investment
spike-based evidence by examining the evolution of the return- AVOL relation
around SEO announcements and around spikes in new issuance activity in gen
eral. According to real options theory, the sensitivity of firm values to volatility
is expected to decline following new issues. We construct four subsamples of
firm-years relative to the timing of firms' SEOs. For example, the first sample
consists of firm-months 13 to 24 prior to an SEO event, while the fourth sample
consists of firm-months 13 to 24 after an SEO.6 There are 9,823 SEOs in our
sample. The results of estimating equation (5) using these event-based samples
are presented in Panel B of Table V.
The evolution of the return-A VOL relation around SEO announcements is
striking. Similar to the investment spike-based results, the coefficients on
A VOL are increasing from event year -2 to year -1. After that, they drop by
approximately 90%, from 1.87 in event year -1 to a statistically insignificant
0.21 in event year 1. In other words, a two standard deviation shock to return
volatility would lead to an almost 7% monthly return in the last year prio
to an SEO on average, while it would lead to a 0.8% return in the first pos
SEO year. The disappearance of the positive relation between returns and
changes in volatility following SEOs is consistent with the hypothesis that
SEO events coincide with exercises of real options, and with the sensitivity of
the value of real options to volatility driving the return- AVOL relation. Similar
to the results of investment spike based tests, the return-A VOL relation start
strengthening again from year 1 to year 2 post-SEO.
6 Similar to investment spikes, when examining the return-AVOL relation around SEOs, we
exclude firm-months within 2 years of at least two SEO events.
Similar to the investment spikebased tests, we want to ensure that the

SEO-based results are not caused by a run-up prior to SEOs and low returns
following SEOs (e.g., Loughran and Ritter (1995) and Ritter (2003)). We there
fore perform a matching procedure similar to that outlined above, while limit
ing the set of potential matches to firms that have not issued seasoned equity
within the previous 3 years. The evolution of the return- AVOL relation around
SEOs for matched firms has a shape similar to that of SEO firms, but the
year-to-year changes in the A VOL coefficients are much smaller than within
the sample of SEO firms. Specifically, the coefficients drop from 1.32 in the
last pre-SEO year to 1.17 in the first post-SEO year. The difference between
the changes in the A VOL coefficients across SEO firms and matched firms is
highly statistically significant.
External financing can take various forms, seasoned equity offerings being
just one of them. Accordingly, we supplement our analysis of the evolution of
the return-A VOL relation around SEO events by looking at broader financing
spikes, defined as firm-years in which the combination of net new issues of
equity and debt exceed 10% of beginning-of-year book assets. There are 60,128
financing spike years in our sample. Panel C of Table V reports the results of
this analysis. Similar to the SEO-based evidence, the sensitivity of firm values
to changes in volatility decreases around years in which firms experience spikes
in their overall financing activity. The coefficient on the change in volatility
drops by more than half from 1 year prior to the financing spike to the first post
spike year. The differences between the changes in the return-AVOL relation
around financing spikes for firms experiencing the spikes and matched firms
are large and statistically significant.
Overall, the evolution of the relation between returns and contemporaneous
changes in volatility around investment and financing spikes is consistent with
the hypothesis that the exercise of real options reduces the proportions of real
options in firm values, leading to a reduction in the sensitivity of firm values
to changes in volatility.
V. Industry Analysis of the Return-Volatility Relation

A. Real Option-Intensive Industries
An additional way of examining whether the positive relation between firm

level returns and contemporaneous changes in volatility can be partially due
to real options whose values are increasing in volatility is to compare the
return-AVOL relation within industries in which real options are more likely
to constitute a larger proportion of firm value to the return- AVOL relation
within industries in which a larger proportion of firm value is attributable to
assets in place.
Using theoretical and empirical studies as guides to identify industries with
plenty of real options, we examine the effect of volatility on returns in the
following industries:
(a) Natural resources industries. Due to the nature of their products and pro
duction processes, natural resources firms are known for their ability to
defer investment, expand, contract, shut down, and restart operations ac

cording to market conditions (e.g., Brennan and Schwartz (1985)). Con
sistent with this argument, Paddock, Siegel, and Smith (1988), Moel and
Tufano (2002), and Fan and Zhu (2010) empirically show that real options
are important for these firms.
(b) High-tech industries and pharmaceutical and biotechnology industries.
Firms in these industries are known for making staged investments, which
allows them to abandon or scale up projects at different points in time
(e.g., Majd and Pindyck (1987), Ottoo (1998), Bollen (1999), and Joos and
Zhdanov (2008)).
We define Fama and French (1997) industries 27 (precious metals), 28 (min

ing), and 30 (oil and natural gas) as natural resources industries. We define
Fama and French industries 22 (electrical equipment), 32 (telecommunica
tions), 35 (computers), 36 (computer software), 37 (electronic equipment), and
38 (measuring and control equipment) as high-tech industries.7 Finally, we
define Fama and French industries 12 (medical equipment) and 13 (pharma
ceutical products) as pharmaceutical and biotechnology industries.
In Table VI we estimate cross-sectional Fama-MacBeth regressions as in
equation (5) for firms operating in high-tech industries, natural resources
industries, and pharmaceutical and biotech industries, as defined above,
and contrast the return-A VOL relations in these industries with the return
AVOL relation in all other industries. In particular, we construct natural
resources/high-tech/pharmaceutical and biotech indicator variables and aug
ment the regression in equation (5) by interacting these indicator variables
with month-to-month changes in return volatility.
If real options are at least partially responsible for the positive relation be
tween returns and changes in volatility, we would expect positive coefficients on
each of these interaction variables. The results, reported in Table VI, are consis
tent with the real options explanation for the positive return-AVOL relation.
The return-AVOL relation is significantly stronger in real option-intensive
industries than in other industries: the mean coefficients on the interaction
variables (0.38 for high-tech, 0.40 for natural resources, and 0.29 for pharma
ceuticals and biotech) are all statistically significant and economically sizable
(they constitute 29% to 48% of the coefficient on A VOL).
In the next subsection we concentrate on one industry that is likely to be
characterized by relatively abundant real optionsoil and natural gasand
examine the return-volatility relation using an industry-specific measure of
volatility of an underlying process and an industry-specific proxy for the mix
of real options and assets in place.
B. Oil and Gas Industry

The oil and natural gas industry serves as a convenient laboratory for exam
ining the effects of the mix of real options and assets in place on the sensitivity
7 This classification is broadly consistent with the definition used in Chemmanur, He, and Nandy
(2010), who define high-tech industries based on three-digit SIC codes.
Table VI
Returns, Contemporaneous Changes in Volatility, and Industry-Wide

Real Option Indicators
This table presents cross-sectional regressions of excess returns on the market factor, beginning
of-year log book-to-market ratio, log market equity, 6-month lagged return for months -7
to -2 relative to the observation month, monthly trading volume normalized by the number of
shares outstanding, month-to-month change in firm-level volatility, AVOLi t, and three interac
tion variables, defined as the product of AVOLi t and an indicator variable equal to one if a firm
belongs to a high-tech industry, natural resources industry, and pharmaceutical or biotechnology
industry, respectively. The sample period is 7/1963 to 12/2008. We define high-tech industries as
Fama-French (1997) industries 22 (electrical equipment), 32 (telecommunications), 35 (computers),
36 (computer software), 37 (electronic equipment), and 38 (measuring and control equipment). We
define natural resources industries as Fama-French industries 27 (precious metals), 28 (mining),
and 30 (oil and natural gas). We define pharmaceutical and biotechnology industries as Fama
French industries 12 (medical equipment) and 13 (pharmaceutical products). We estimate the
regressions monthly and report time-series means of coefficient estimates along with i-statistics
obtained using Newey-West autocorrelation- and heteroskedasticity-consistent standard errors
of monthly coefficient estimates in parentheses. R2 refers to the average monthly ft2. We also
present the impact of large AVOLi t on returns for various industry groups. Large AVOLit refers
to AVOLi t that is two standard deviations above the mean; such change equals 3.66%.
Market 0.108
(3.09)
Log (B/M) 0.715
(8.89)
Log (equity value) -0.131
(-2.81)
Lag (6-month return) -0.001
(-0.58)
Volume 2.751
(11.26)
A Volatility 0.826
(7.08)
A Volatility * high tech 0.383
(4.25)
A Volatility * natural resources 0.400
(4.01)
A Volatility * pharmaceuticals 0.242
(2.35)
R2 0.110
# Months 539
Impact of large AVOLl t on return
High tech 4.43%
Natural resources 4.49%
High tech 3.91%
High tech 3.02%
of firm value to changes in volatility of a process underlying its real options.

First, unlike in the large-sample tests above, in which we use volatility of stock
returns as a proxy for the volatility of the process underlying real options, a
more direct proxy is available for oil and gas firms: the volatility of relative oil
price changes. Second, instead of relying on various indirect proxies for real
options, we are able to construct more direct measures of the proportion of r

options in oil and gas firm values. These measures are based on proporti
of developed oil and gas reserves out of total proven reserves. The larger
proportion of a firm's remaining undeveloped reserves, the more real option
is expected to have because undeveloped reserves are unexercised real options
In our sample of 72 firms over the period 1995 to 2009, the mean proportion
developed oil and gas reserves is 27%. More importantly, there is substan
cross-sectional variation in the proportion of developed oil (gas) reserves: the
standard deviation is 21% (20%).
To examine the effect of the proportion of undeveloped reserves on the return
A VOL relation, we estimate the following regression:
Vit r f,t = of | /?A OILVOLt -1-1 j PR 0 Piyt A Oil j \() /> < -(- OOILRETt -I- y ^\u\r: t
+ 7 xiit +i,t, (12)

where A OILVOLt is the month-to-m
PRO Pit AOILVOLt is an interaction var
tion of undeveloped reserves, PROP^, a
volatility of daily relative oil price ch
takes one of three values: the proporti
portion of undeveloped gas reserves,
proportions, proxying for the overall p
compute OILVOLt as the standard de
in the price of Brent Crude oil durin
relative change in oil price (i.e., oil re
in equation (3). We estimate equation
(2009), cluster standard errors by mo
presented in Panel A of Table VII.
Not surprisingly, an increase (decrea
crease (reduction) in oil firm value. Sim
value is positively related to contempor
Importantly, the coefficients on the in
turn volatility and all three measures o
highly significant. An increase (decr
positive (negative) impact on the value
to develop a larger proportion of the
standard deviation increase in the pr
sociated with a 0.4 to 0.5 increase in
Stated differently, the relation betwee
return volatility is 224% to 375% stron
veloped reserves that is two standard
for firms with no undeveloped reserves
calendar-time and event-time evidence
8 We convert cubic feet of natural gas to a barr

by 0.0001767.
Table VII
Returns of Oil and Gas Firms and Contemporaneous Changes

in Volatility of Oil Returns
This table presents panel regressions of oil and gas firms' excess returns on the market fac
tor, beginning-of-year log book-to-market ratio, log market equity, 6-month lagged return for
number of shares outstanding, monthly relative change in oil price (monthly oil return), month-to
month change in oil price volatility, and interaction variables equal to the product of the proportion
of undeveloped oil/gas/total reserves and month-to-month change in oil price volatility. The sample
period is 1/1995 to 12/2009. Oil and gas firms are those belonging to Fama-French (1997) indus
try 30. Relative oil price changes and their volatilities are computed using Brent Crude oil daily
prices. See Section VI for a discussion of developed and undeveloped reserves. Standard errors are
clustered by month. We also present a summary of the economic effects of the proportion of unde
veloped reserves on the AOILVOLi,(-return relation. No (high) undeveloped reserves means that
the effect of AOILVOLi t on returns is calculated using the value of the proxy for the proportion of
undeveloped reserves at zero (two standard deviations above the mean). Large A OILVOLi t refers
to AOILVOLi t that is two standard deviations above the mean; such change equals 1.95%.
Type of undeveloped reserves
Oil Gas Total
Market 0.096 0.060 0.059 0.065

(0.97) (0.61) (0.60) (0.66)
Log (B/M) 1.210 1.236 1.224 1.211
(2.92) (3.00) (2.96) (2.93)
Log (equity value) -0.676 -0.643 -0.656 -0.655
(-7.64) (-7.29) (-7.43) (-7.42)
Lag (6-month return) 0.010 0.010 0.010 0.010
(2.39) (2.41) (2.48) (2.43)
Volume 1.857 1.735 1.783 1.784
(12.33) (11.53) (11.84) (11.84)
Relative oil price change 0.427 0.425 0.427 0.427
(22.78) (22.79) (22.84) (22.83)
A Oil volatility 0.652 0.489 0.553 0.559
(3.38) (2.53) (2.86) (2.90)
Prop, undev. reserves * A oil volatility 2.659
(8.70)
(6.79)
(6.30)
R2 0.081 0.089 0.086 0.085
# Observations 8,192 8,192 8,192 8,192
Total effect of AOILVOLl t on return
High undev. reserves 2.324 2.075 1.809
No undev. reserves 0.489 0.553 0.559
Impact of large AOILVOLn on return
High undev. reserves 4.53% 4.04% 3.53%
No undev. reserves 0.95% 1.08% 1.09%
real options and assets in place affects the sensitivity of firm value to changes
in the volatility of its underlying assets.
VI. Real Options and Asset Pricing Tests
In this section we present results of a test of the real options explanation for
the positive relation between returns and changes in volatility that is based
on the performance of asset pricing models for real options-based firms and
assets in place-based ones. Da, Guo, and Jagannathan (2012) argue that real
options may be one of the reasons for the poor performance of the CAPM in
explaining the cross section of returns. The idea is that, since firms consist
of multiple investment options that may be exercised at different times, the
CAPM (or more generally, any asset pricing model that does not account for
real options) may be unsuccessful in explaining firms' stock returns even if
it explains returns to individual projects perfectly (e.g., McDonald and Siegel
(1985) and Berk, Green, and Naik (1999)). This logic implies that the success of
an asset pricing model should be related to the proportion of real options in the
value of firms whose returns are used in testing the model. In this section we
examine the ability of the CAPM and the Fama and French (1993) three-factor
model to explain returns of firms with different mixes of real options and assets
in place.
All of the results in the previous four sections show that the sensitivity of
returns to changes in volatility is positively associated with various measures
of real options. This implies that sorting firms by the sensitivity of their returns
to changes in volatility would result in groups of firms with varying mixes of
real options and assets in place. It would then be possible to compare the
performance of asset pricing models across real option-based subsamples of
firms.
To estimate the sensitivity of firm value to volatility, for each firm i in month
t we estimate the following firm-level time-series regression using data during
months r e(t 60, t 1):
Tf,t &i,t m,z ~ T*f ,t) ~t" Yi.t z i x. (13)
The estimated coefficient on AVOLia, y^t, is the sensitivity of firm z's value
to the change in volatility assigned to firm i in month t. Each month we sort
firms by yTt and assign them into real options quintiles. The highest y quintile
contains firms with the largest proportion of their value represented by real
options, while the lowest y quintile contains assets in place-based firms. Then,
within each real options quintile each month we sort firms by their estimated
market beta, $)t , and assign them into five beta quintiles.
We compute monthly value-weighted returns of the resulting 25 portfolios
and estimate time-series regressions of portfolio excess returns on the excess
returns of the market portfolio:
7*p,t f f.t "I" Ppi^m,t I"f,t) "I" p,t-> (14)
where rpJ is the monthly return of each of the 25 portfoli

VIII presents the estimated intercepts of equation (14) for ea
folios as well as the loadings on the excess market return. In
each real options group, we compute the average pricing err
absolute value of the intercepts of equation (14)) and the Gibbons, Ross, and
Shanken (1989) statistic (GRS) for testing whether the pricing errors of the
five beta portfolios within each real options quintile are jointly zero. The mean
pricing error for the lowest real options group is 14 basis points per month, sub
stantially lower than the mean pricing error for the highest real options group,
which equals 25 basis points per month. The GRS .F-statistic is a statistically
insignificant 1.32 (p-value of 0.25) for the lowest real options quintile, while
it is a highly significant 5.4 (p-value of less than 0.001) for the highest-real
options quintile. With the exception of the second quintile, the GRS statistic is
monotonically increasing as we move from portfolios of firms with more assets
in place to portfolios of firms with more real options.
The evidence in Panel A of Table VIII shows that, while the CAPM is com
fortably rejected when the testing portfolios include firms with abundant real
options, it performs much better when portfolios of assets in place-based
firms are being used. This is consistent with the argument in Da, Guo, and
Jagannathan (2012) and with the models of McDonald and Siegel (1985) and
Berk, Green, and Naik (1999).
In Panel B of Table VIII we perform similar tests of the Fama and French
(1993) three-factor model. Specifically, we estimate time-series regressions of
the form
rp.t r f,t 01 p + Pi,p(rm,t rf,t) + Pi.pHMLt + fopSMBt + eij (15)
for the 25 portfolios discussed above. The results are quite similar to the CAPM
based evidence. There is a large difference between the average pricing error
within the lowest real options-based group and that within the highest real
options group, 0.08 and 0.21, respectively. In addition, the three-factor model
can be comfortably rejected for real-option-intensive firms CF-statistic = 4.07,
p-value = 0.001), while it cannot be rejected for assets in place-based firms
CP-statistic = 0.58, p-value = 0.72).
Overall, the evidence in this section suggests that nonlinearities of firm
value with respect to the value of firms' projects/investments could contribute
to failures of asset pricing models in explaining returns of certain portfolios.
The evidence also suggests that samples of firms with relatively low sensitivity
of value to volatility (i.e., firms with relatively few real options) may be better
suited for testing asset pricing models, consistent with the return-volatility
relation being correlated with the mix of real options and assets in place.
VII. Aggregate Returns and Aggregate Volatility

As mentioned in the introduction, numerous studies (e.g., French, Schwert,
and Stambaugh (1987), and Duffee (1995)) find a negative contemporaneous
12 a 03
3 -3 1 G %u lo CD 00 to o cu
a -So CO o t> LO
T3 P H TH CN rH rH CN
S
^g 03
s 'S S
^ ft
o o o o o
00 - Jh
o CO o3
o ft
03
S3 GO

O 03 <N O Gi cn O
XJ 6 8 '* PS CO CO I> I>
<d a o rH CO l CN id
t, fe cc 2.
0 <u
CO a)
CD -O CO oo rH LO rH CO rH oo
CO ^
.2 a 2 o lo
q 00 CO lo CO rH 00
rH CO <n 00 rH ai CN ai Tp CN
t> ca
<2 T3 lo tO LO
J^
LO lo
.W 03
1 T3 "S "
- P
lO _ 'C S h CO
rH
CO
rH t>
rH
CN
rH
CO
i>
O
o
LO CN
CO lo
lo
CO
IN a> _d
rH CO CN lo CN CO lo LO LO
<*I ^
i *n v &< S o
1
rH o
1
rH o
1
rH o
1
rH o rH
cc ^ _u
91 S .S
g.s fi
o3 -S ^3
3 . rH 00 rH CO LO CN LO CO LO Gi
S3 03 CO 0} o tO rH 00 LO O
81 :l.s Tji <N t> d 00 o rH Gi

rH CO
s' N'
| s | 1 CO
8 S.
w
X -M !o
i 03 t2
^1 i 3 ~
v. a rH
co
LO
Gi
t>
t>
co
CO
D
00
co
Gi LO
CO
CO
Gi
O
LO
CO
II "S > 03
O -t-s
O a) 0)
<n
o
<N
rH
CM
o rH
<N o
o
rH
rH
o
o
rH
rH
(N
o
rH
rH
be 0J 1 1 1
a> r2 Û"3Q)^
cd O >
s
CD CO rr<
o 'S a; co LO Gi CO CO CO CO (N rH
e rH LO 00 <N CO o 00 <N CO
f> 3 2 s5 o CO o rH 00 rH CO o rH
c 1
CO CO CO CO 1 CO
>5 "S W W W
* 5
5!
+3 co <K >H u
cJ 10 O
s
S o ^ -S H <N
(M
CO t>
o
rH
O
CO
LO <N
CO CO o t>
x to LO CO rH
03
o o rH Gi rH Gi rH Gi O Gi
0)
co a3 be
o 03 a o rH o o O o o o o O
Table VII 03 .
w T3 .S a 1
CO 5 0) 03 o Ol,
r C3 c
ft
As et Pric ng Tests CO cn CO CO (N Gi lo CO
a>
- JJ CO CO rH CO rH T}H rH (N co
cd tr +> o co rH ai rH <N rH ai CN CO
ft O H v' s' 1' s' s'
<N <N CO (N (N
fc ^
03 73
?! 00
GO
o
oo
LO
l>
t>
CO
rH
Gi
CO
I>
CO
CN
rH
l>
o
oo
lo
CO
rH
I>
lo
I>
t>
LO
CO t>
I>
CO
ft *c bp o o O o o o O d o o
2
u&

c t
a g 2 S to S> C5 P 00 CO P Gi
o .g
B Jh lo <N <N t> CO rH lo i> CO
2? w
03 03 +3 03
c2 rH CO c4 rH
1'
rH
oj lo o 00
<N CN CO CN CN
2 "
gg 03
o
t-H
rH
C
o CO CN
TP t>
O
00
rH
CO
rH
CO o
CO
CO
c CO <N rH D <N CO rH 00
o
CM
q
PQ d
? CO o o o o o o o o o
d 2
IIIcu
03 yH
s1 a JJ <*
38
c 3 rH CN CO LO
Ch T3
03 r*
(S'SI
CO n
fli tvJ a>
! a) 3"
> f
> c o ft Oh &tj Oh Go Oh ft Caj
1 ** S
; <
<D 3 Vu 3
u Ph K
fn tH M Ph
s s
Sh tH
8
Vh Ph
8
t-H P-l
1! 05
8.gS a ^ | -2 s c w^ PhW^ "a ^
a d2 b S

be
Cq o rH CO 00
Ch rj Jh
co .5 o 00 rH CTi o
0) o H o rH o O cq
SI s o O o o d
CO 00 rH CO rH t>
m 00 o cq o
O o rH rH rH
i> oo Cq Cq C5 LO 00 rH CO OJ P CD CO in
in 00 CO in
T1 (N lO 05 (N CO 00 i>
q rH CO xjn
o 00 00 id o o 00 rH rH tjh' <N cq cq* in i> CO in
rH m rH
S
rH rH
s 1 1 1
rH
w w
rH a o CO (N o> 05 rH CD I> G5 O [> CO CO m o rH

CI o 00 rH O Cq (N rH CO 00 lO CO m rH rH 05 CD o G5
O CO CD rH CO CO m rH CO rH cq CO cq CD CO cq m
rt|
o rH o d O rH o o O rH o o o rH o o o rH o o
1 1 1 1 1 1 1 I 1
cq iH rH 05 OS P cq rH <N 00 CO CD cq 00 CO CO
cq CD CO CO in CD <N in rH 00 rH CD in
cq CO 00 in
q
cq o in rH CD cq lO o in rH o 00 CO cq
rH tH CO lO in
jo
I S 1 1 S 1 s S CO
tH lO 00 T1 i> CD rH 00 in rH cq rH Gi o rH Gi rH t> cq Ci
W r-1 rH 00 os rH i> CD o> tH 00 CD 00 rH o in D rH
rH <N rH CM cq rH O tH rH rH rH O rH cq O rH rH tH rH
'd o rH o o o rH o o o rH o O O rH o o o rH o o
o
1 1 1 1 1 1
s
00 Cq rH CO 00 CD rH 00 CO 00 CO CO CD CO Gi rH 00 CO
cd lO CO CD 00 rH 03 rH 00 rH CO rH cq o Tt< in CO
cq
o t> CO o o t> CD CO o d JO rH o d in rH rH CO CD rH
CO CO CO CO CO
cq o cq CO o CO t m O 00 m CD rH o o CD
oi CD CO CM Oi CO (M cq cq rH rH O LO o Oi m
o o rH o o 05 cq rH O o cq o O o Cq o rH o cq o
a o rH o o o o o d o rH o o o rH o o o rH o o
a>
H
fe
ctf
s 00 CD rH CD CO c^ OJ cq o CO Oi 05 o o CD
O 05 o 00 CO CO I> rH in rH O
a3 CO q q rH
o oo in cq o cq CD cq o in in in o cq in in rH 00 CD O
cq CO CO CO cq
1 v'
1 I
"S
t-WtCNtDNlN^îOO^tOOtOHHiOTfN
CC
^"^Loaiocot-or-iooococoiooaâiâio
CLh
qcoiNqqcoNHqQiNHqcoNHHcqiNq
oooooooooooooooooooo
I I I I I I
0 00 cq CO CO CO CO in Ol 00 00 cq cq rH 00
G5
Oi 00 CO rH cq CO rH Ci o cq CD cq
q t rH c m rH
q
o CO rH rH l> CD o rH i5 cq d d CD CD
5 o t> o CO
cq cq CO cq Cq
00 CD rH CD m CD CO cq CD Oi CO CD 00 t> in cq a>
rH cq I> CO rH CO rH O o o o Oi CD CD rH Th
o 00 o CD cq 00 o
rH t> rH o rH r> cq rH rH rH rH
q rH
o o o o o o o o o o o o d o o o o o o o
a ft, a Ct, pi Clj Oh Cu
8g , 8M, 8 cd !3
sS
feh fe H J n Vh Ch
sgys s ^
S3 !3 d FP sr- h-3 CQ
laag 3w
W m S W C/] -S K M
relation between aggregate market excess return and aggregate market volatil
ity, in contrast to the positive relation at the firm level. In this section we
examine the potential reason for the discrepancy between the firm-level and
aggregate evidence.
We begin our analysis by replicating the existing evidence on the relation
between aggregate volatility and stock returns. The first column in Panel A
of Table IX presents estimates of the regression of market excess return on
contemporaneous market volatility:
rm,t - r-fj = a + pVOL^t + et, (16)
where rm t is the return on the value-weighted market portfolio in month

is the risk-free rate in month t, and VOLm t is the standard deviatio
returns on the value-weighted market portfolio during month t, esti
in equation (1). Similar to the evidence in past studies, the relatio
market excess return and contemporaneous aggregate volatility is negative
and highly statistically significant.
In the second column of Panel A we estimate a regression similar to equa
tion (16), but instead of return volatilities we use changes in volatility, defined
in equation (2):
rm,t ~~ rf,t a + P AVOLmj + et. (17)

As in the case of the levels of aggregate volatility, market returns are neg
atively and significantly related to contemporaneous changes in aggregate
volatility.
As mentioned in the introduction, various hypotheses, such as the leverage
effect and increased risk premia caused by increased volatility, have been pro
posed as explanations for the negative relation between market returns and
(changes in) aggregate volatility. One additional reason for the negative rela
tion between returns and return volatility at the aggregate level is a variable
potentially omitted from equations (16) and (17) that could affect both (levels
of and changes in) aggregate volatility and returns on the market portfolio.
For instance, an improvement (deterioration) in aggregate market conditions
could have both a positive (negative) impact on market return and a negative
(positive) effect on aggregate volatility. In other words, the negative relation
between aggregate market returns and aggregate volatility could potentially
be spurious and driven by an omitted proxy for contemporaneous changes in
macroeconomic conditions.
To examine whether the omitted variable problem may be responsible for the
negative relation between aggregate returns and return volatility, in Panel B
of Table IX we estimate the relation between individual stock returns and con
temporaneous changes in aggregate volatility, while controlling for the returns
on the Fama and French (1993) three factors:
n,t ~ rfit = at + PiAVOLm,t + yu(rm,t - rftt) + Y2,iSMBt + Ys,iHMLt + (18)
Table IX
Time Series Regressions

Panel A presents regressions of aggregate (value-weighted) monthly excess return on aggregate
volatility and month-to-month change in aggregate volatility. Aggregate volatility and change
in it refers to the volatility of the value-weighted monthly return on all listed stocks. Panel B
presents regressions of firm-level excess returns on changes in aggregate volatility and the Fama
and French (1993) three factor returns (MKTRF, HML, and SMB). Panel C presents regressions
of firm-level excess returns on changes in aggregate volatility and the Fama and French (1993)
three factor returns for subsamples of firms grouped by measures of investment opportunities
and by measures of flexibility. Only the coefficients on the change in volatility are reported. The
last column reports the difference between the estimated coefficients on the change in aggregate
volatility between the extreme real options groups. The sample period is 7/1963 to 12/2008. t
statistics are reported in parentheses. In Panel A, ^-statistics are obtained using Newey-West
autocorrelation- and heteroskedasticity-consistent standard errors. We estimate panel regressions
in Panels B and C and cluster standard errors by month.
Panel A: Dependent Variable: Aggregate Excess Return
Aggregate volatility 3.06

(-8.57)
A Aggregate volatility 3.88
(-8.65)
R2 0.118 0.122
# Observations 540 539
Panel B: Dependent
A -6.87
Aggregate -2.33 0.01
volatility
(-3.36) (-1.02) (0.00)
MKTRF 1.18 1.08
(8.93) (5.67)
HML 0.23
(0.98)
SMB 1.27
(6.12)
R2 0.010 0.040 0.058
# Observations 3,040,952 3,040,952 3,040,952
Panel C: Dependent Variable: Firm-Level Excess Return. By Real Option Groups
Ranking 1 2 3 4 5 5-1
Book assets
A Aggregate volatility 0.460 0.205 -0.037 -0.250 -0.175 -0.635
(4.94) (4.65) (-1.05) (-6.84) (-7.50) (-6.61)
Age
A Aggregate volatility 0.226 0.564 0.137 0.005 -0.244 -0.470
(3.15) (8.71) (2.49) (0.14) (-10.37) (-6.22)
RD/assets
A Aggregate volatility 0.004 0.405 0.400 0.809 0.719 0.715
(0.05) (4.06) (3.98) (5.82) (4.05) (3.67)
Future sales growth
A Aggregate volatility 0.117 0.000 0.085 0.355 0.672 0.555
(1.65) (0.01) (1.56) (5.77) (8.25) (5.14)
(continued)
Table IX(Continued)
Panel C: Dependent Variable: Firm-Level Excess Return. By Real Option Groups
Ranking 1 2 3 4 5 5-1
Earnings convexity
0.095
A Aggregate volatility -0.086 -0.082 0.393 0.349 0.254
(2.25) (-1.87) (-1.50) (6.19) (4.23) (2.74)
Sales convexity
0.113
A Aggregate volatility -0.044 0.105 0.013 0.265 0.152
(2.86) (-1.10) (1.91) (0.22) (2.52) (1.35)
Union membership
0.141
A Aggregate volatility 0.496 0.312 -0.683 -0.509 -0.650
(2.24) (3.24) (2.11) (-6.94) (-4.05) (-4.62)
Using individual stock returns instead of aggregate market returns as the

dependent variable allows us to simultaneously control for market factors as
well as aggregate volatility. We use a panel data approach to estimate the
parameters of equation (18) and, following Petersen (2009), cluster standard
errors by month.
As is evident from the first column in Panel B, in which aggregate factor
returns are excluded from equation (18), the coefficient on the change in ag
gregate volatility is negative and highly significant, similar to the aggregate
evidence in Panel A. Augmenting the regressions by the contemporaneous re
turn on the market portfolio, in the second column, reduces the magnitude of
the coefficient on the change in market volatility by about two-thirds, but it
remains negative and significant. Further augmenting the regressions by re
turns on the SMB and HML factor mimicking portfolios, in the third column,
results in zero relation between returns and changes in aggregate volatility.
Finally, in Panel C we investigate the relation between firm-level returns
and aggregate volatility changes by examining the coefficient on the change
on aggregate volatility across different portfolios based on proxies for growth
options and proxies for operating flexibility. To conserve space, we only report
the coefficients on A VOLm t. Consistent with the theory, we find that aggregate
volatility has a positive effect on the value of real options-based firms and
a negative effect on the value of assets in place-based firms. Specifically, we
find that small firms, young firms, high R&D firms, high growth firms, and
more flexible firms have a strong positive aggregate volatility-return relation
while large firms, old firms, low R&D firms, low growth firms, and less flexible
firms tend to have a much weaker and sometimes negative relation between
returns and changes in aggregate volatility. This result is important, as it
demonstrates that the negative relation between aggregate market returns
and contemporaneous volatility is not necessarily inconsistent with the positive
relation at the firm level. Further, it shows that real options are important even
at the aggregate level.
VIII. Robustness Checks
In this section we examine the robustness of our results. First, we exam

ine whether month-to-month changes in daily return volatilities are driven
by transitory jumps in stock prices as opposed to permanent changes in the
diffusion component of stock price evolution. We then proceed to examine pos
sible nonlinearities in the relation between firm value and changes in volatility
and the potential joint effects of shocks to volatility and shocks to liquidity on
returns. Next, we examine whether the "leverage hypothesis" or the "resale
option hypothesis" can potentially explain some of the results. Finally, we ex
amine the robustness of our results involving oil and gas firms to controlling
for expected changes in oil return volatility. In this section, to save space, we
do not tabulate the results but instead discuss the main findings. The results
that we discuss below are available in the Internet Appendix.9
A. Are Changes in Return Volatility Driven by Jumps in Daily Stock Prices?
Throughout the paper we use the volatility of daily stock returns as a proxy
for the volatility of the process underlying firm value. However, an alternative
interpretation of month-to-month changes in daily stock return volatility is
that they mainly reflect transitory jumps in daily stock prices. If high return
volatility in a given month is driven by large daily returns that occurred during
that month, then we could observe a positive relation between monthly returns
and return volatility.
In this subsection we discuss a battery of tests designed to distinguish be
tween the two interpretations discussed above. We begin by computing the
correlations among return volatility, VOLit, month-to-month change in volatil
ity, AVOLi t, maximum daily return within a month, MAXl t, and minimum
daily return within the month, MINij. If the positive return-volatility rela
tion is driven by positive daily price jumps, we should expect to find a higher
(absolute) correlation between AVOLn and MAXl t than between AVOLi t and
MINi t. However, the former correlation is 0.338, while the latter one is -0.354,
inconsistent with the hypothesis that the positive return- A VOL relation is
driven primarily by positive daily price jumps.
Second, if changes in return volatility are driven entirely by transitory price
jumps, we would expect changes in volatility to reverse completely. To examine
whether this is the case, we analyze the evolution of volatility before and after
large volatility shocks. Specifically, we track volatility for 12 months before and
after firm-months in which the change in volatility is in the top or bottom 10%
of the sample. Close to half of positive and negative changes in volatility seem
to be permanent, in the sense that they are not reversed within 12 months.
Third, to ensure that our results are not driven by positive price jumps, we
estimate the return-A VOL relation while omitting potential jumps from the
sample. Specifically, we replicate the cross-sectional tests in Tables III and IV
9 The Internet Appendix is available on the Journal of Finance website at http://www.

afaj of.org/Supplements.asp.
while excluding the top 5% and bottom 5% of daily returns, and obtain resu
similar to those in the body of the paper. The Internet Appendix reports th
results obtained using the middle 90% of return observations.
Fourth, Frazzini and Lamont (2007) show that returns around earnings
announcements are typically positive, and trading volume around earnings
announcements is abnormally high. If abnormal volume is associated with ab
normal return volatility, then the positive return- A VOL relation may be driven
by the effect of earnings announcements. To ensure that this is not the case,
we repeat the cross-sectional tests while excluding earnings announcement
firm-months and obtain results consistent with those in Tables III and IV.
Fifth, we remove the potential mechanical relation between returns and
return volatility by computing them on different days. Specifically, we com
pute monthly returns while using returns on even days, and compute return
volatility using returns on odd days. If the return- AVOL relation is driven by
jumps in daily stock prices, we should not see any relation between returns
and return volatility when they are computed on different days. However, the
cross-sectional results stay intact, suggesting that they are not driven by daily
stock price jumps.
Finally, we repeat the tests while extracting the diffusion component of
stock return volatility. In particular, we follow the methodology proposed by
Barndorff-Nielsen and Shephard (2004, 2006) and Andersen, Bollerslev, and
Diebold (2007) to separate the continuous sample path variation from the dis
continuous jump part of the variation using the realized bipower variation
measure. We then repeat the cross-sectional tests while using the estimates of
the diffusion component of the price process instead of raw price changes and
obtain results similar to those reported in the body of the paper.
Overall, the tests discussed in this subsection suggest that the relation be
tween proxies for the mix of real options and assets in place and the sensitivity
of returns to changes in volatility are not driven by positive daily price jumps.
B. Nonlinearities in the Return-AVOL Relation
To verify that potential nonlinearities in the return-A VOL relation do not

drive our results, we estimate regressions in which, instead of interacting
A VOL with measures of real options directly, we use a four-by-one vector of
interaction variables, defined as the product of the contemporaneous change in
return volatility and a dummy variable equal to one if the firm belongs to the
second (third, fourth, fifth) real options quintile, and equal to zero otherwise.
The results indicate that, in the majority of specifications, the return-AVOL
relation is monotonically increasing as we move from the lowest real options
quintile to the highest real options quintile.
C. The Joint Effects of Volatility and Liquidity
Bandi, Moise, and Russell (2008) show that shocks to market liquidity an
shocks to market volatility jointly affect market returns. To investigate the join
effects of changes in firm-level volatility and firm-level liquidity on returns, w
augment the regressions in equation (5) by including month-t

in Amihud's (2002) illiquidity measure. Including this meas
explanatory variables does not change the coefficients on the
ity substantially. The coefficients on illiquidity innovation ar
specifications, consistent with increases (decreases) in illiq
contemporaneous drops (increases) in stock prices.
D. Leverage Hypothesis
A potential concern is that there could be sizeable difference
ity across the various subsamples in split-sample tests. We add
as follows. We first split the sample each month into deciles of
split each of the 10 subsamples into quintiles of volatility
the return-volatility relation across leverage subsamples while
volatility reveals that generally there is no particular patte
between leverage and the sensitivity of returns to volatility.
E. Resale Option Hypothesis

Scheinkman and Xiong (2003) develop a theoretical model in
fidence generates divergence of opinion among investors abou
of an asset. They show that, in the presence of short-selling c
heterogeneous beliefs create a situation in which a buyer o
an implicit option to resell the asset at a higher price to a mo
investor. Because the value of this resale option is increasing i
of the fundamental process, the price bubble generated by
is positively correlated with volatility. Therefore, the resale o
provides an alternative explanation for the positive relation
in volatility and stock returns documented by Duffee (1995).
Although Battalio and Schultz (2006), Schultz (2008), and Grif
do not find evidence supporting some of the main predictions o
hypothesis, in this subsection we examine whether this hypot
the positive volatility-return relation at the firm level. As
Scheinkman and Xiong (2003), the sensitivity of the price bub
the volatility of the underlying asset increases as the level of
increases. Thus, if the resale option hypothesis is partially
our results, then the positive relation between changes in vol
returns should be stronger in times when investors' confi
high. To test this prediction, we use Baker and Wurgler's
investor sentiment to proxy for the level of overconfidence in
proxy seems to be a natural measure to test the resale opt
cause it tends to be high during episodes that have been classi
bubbles" and it has been shown to capture overvaluation in th
stock returns (see Baker and Wurgler (2006)).
To examine the effect of overconfidence on the volatility-re
augment the regression in equation (5) by including an add
which we interact the measure of investor sentiment with the change in volatil
ity. Contrary to the predictions of the resale option hypothesis, we do not find
any evidence that the sensitivity of stock prices to changes in volatility in
creases during periods of high investor sentiment. This evidence points to an
alternative interpretation of the investor sentiment measureit could be that
investors are optimistic and enthusiastic in those years, not because of any ir
rational exuberance, but because the economy is doing well. Therefore, because
firms are more likely to exercise their real options during good economic times,
the sensitivity of stock prices to changes in volatility declines during those
periods. Overall, our results do not support the idea that the resale option
drives the observed positive correlation between stock returns and changes in
volatility.
F. Controlling for Expected Changes in Volatility of Oil Returns

Carlson, Khokher, and Titman (2007) show theoretically and empirically
that future volatility of oil returns is higher when the forward curve is steeply
upward sloping (in contango) or downward sloping (backwardated). Thus,
it could be argued that changes in oil return volatility used in the tests in
Section V.B may be partially predictable. To address this concern, we follow
Carlson, Khokher, and Titman (2007) and estimate the unexpected component
of changes in oil return volatility. Specifically, we first de-seasonalize oil
return volatility by regressing it on 12 monthly dummy variables. We then
regress de-seasonalized oil return volatility on the de-seasonalized slope of
the term structure of oil futures, obtained from Tick data, and its square.10
We next compute month-to-month unexpected changes in oil return volatility
as the month-to-month differences in the residuals of this regression. Using
the unexpected component of the change in oil return volatility instead
of the raw change in oil return volatility does not affect the qualitative
results.
IX. Conclusions
In this paper we provide evidence suggesting that an important reason

positive contemporaneous relation between firm-level returns and firm
volatility, first documented by Duffee (1995), is the presence of real o
Consistent with the theoretical argument that the value of a real option is
increasing in the volatility of the underlying process, we find that firms with
more real options exhibit a stronger sensitivity of firm value to underlying
volatility.
Using various proxies for the proportion of firm value represented by invest
ment opportunities, we show that the value of firms with abundant invest
ment opportunities is highly sensitive to changes in volatility, while firms that
10 As in Carlson, Khokher, and Titman (2007), the slope of the term structure of oil futures is
computed using the nearest and third-nearest contracts.
derive most of their values from existing assets exhibit su

sensitivities of values to volatility. In addition, we find that t
relation between volatility and returns is stronger for firms
operating flexibility and more convex value functions. We als
with plenty of growth and strategic options (high-tech firms
and biotechnological firms) and with high levels of operating f
resources firms) tend to exhibit a stronger volatility-return
tion, we focus on the oil and gas industry, for which we are
industry-specific measures of the mix of real options and
of the volatility of the underlying process (oil price). We find
with the large-sample evidence.
Time-series results further reinforce the real options hyp
riods of abnormally high investment activity and abnormally
activity, and in particular SEOs, as proxies for real option
find a highly economically and statistically significant decline
ity of firm values to volatility around times of real option ex
with the hypothesis that a reduction in the amount of remain
reduces the sensitivity of firm value to underlying volatility.
We also build on recent findings on the effects of real optio
mance of asset pricing models. Da, Guo, and Jagannathan (201
the presence of real options, the CAPM cannot explain equity
equity expected returns become nonlinear functions of the ex
of the underlying projects. Thus, if the positive correlation b
and returns is mainly driven by real option effects, then one
performance of the CAPM (or any other asset pricing mod
within subsamples of firms with a relatively weak return-
(i.e., firms with relatively few real options). Consistent w
find that the CAPM and Fama and French (1993) three-fact
better within subsamples of firms with a relatively weak relation between
returns and changes in volatility than within subsamples of firms with a rel
atively strong return-volatility relation. This provides another piece of evi
dence linking the positive return-volatility relation at the firm level to real
options.
Finally, our paper suggests a rational explanation for the discrepancy be
tween the negative volatility-return relation at the aggregate level and the
positive relation at the firm level. We argue that the negative relation be
tween aggregate returns and aggregate volatility could be driven by com
mon underlying economic factors affecting both variables. Consistent with this
argument, we show that, after controlling for aggregate market conditions,
aggregate volatility has no relation with individual stock returns. Moreover,
the relation between firm-level returns and aggregate volatility tends to be
significantly positive for real options-based firms, while it is much weaker and
sometimes negative for assets in place-based firms.
In general, our findings support the real options-based explanation for the
positive relation between volatility and returns. This result sheds light on the
fundamental issue of how volatility affects asset prices.
REFERENCES
Abraham, Katharine, and James Medoff, 1984, Length of service and layoffs in un
work groups, Industrial and Labor Relations Review 38, 87-97.
Albuquerque, Rui, 2012, Skewness in stock returns: Reconciling the evidence o
gregate returns, Review of Financial Studies 25, 1630-1673.
Amihud, Yakov, 2002, Illiquidity and stock returns: Cross-section and time-series
of Financial Markets 5, 31-56.
Andersen, Torben, Tim Bollerslev, and Francis Diebold, 2007, Roughing it up: Inclu
ponents in measuring, modelling and forecasting asset return volatility, Review
and Statistics 89, 701-120.
Anderson, Christopher, and Luis Garcia-Feijoo, 2006, Empirical evidence on ca
growth options, and security prices, Journal of Finance 61, 171-194.
Ang, Andrew, Robert Hodrick, Yuhang Xing, and Xiaoyan Zhang, 2006, The cross-
ity and expected returns, Journal of Finance 61, 259-299.
Ang, Andrew, Robert Hodrick, Yuhang Xing, and Xiaoyan Zhang, 2009, High idiosy
and low returns: International and further U.S. evidence, Journal of Financi
1-23.
Baker, Malcolm, and Jeffrey Wurgler, 2006, Investor sentiment and the cross-section of stock
returns, Journal of Finance 61, 1645-1680.
Bandi, Federico, Claudia Moise, and Jeffrey Russell, 2008, The joint pricing of volatility and
liquidity, Working paper, University of Chicago.
Barber, Brad, and John Lyon, 1997, Detecting long-run abnormal stock returns: The em
pirical power and specification of test statistics, Journal of Financial Economics 43,
341-372.
Barndorff-Nielsen, Ole, and Neil Shephard, 2004, Power and bipower variation with stochasti
volatility and jumps, Journal of Financial Econometrics 2, 1-37.
Barndorff-Nielsen, Ole, and Neil Shephard, 2006, Econometrics of testing for jumps in financi
economics using bipower variation, Journal of Financial Econometrics 4, 1-30.
Battalio, Robert, and Paul Schultz, 2006, Options and the bubble, Journal of Finance 61,
2071-2102.
Berk, Jonathan, Richard Green, and Vasant Naik, 1999, Optimal investment, growth options, a
security returns, Journal of Finance 54, 1553-1607.
Bernard, Victor, and Jacob Thomas, 1989, Post-earnings announcement drift: Delayed pric
sponse or risk premium, Journal of Accounting Research 27, 1-35.
Bernardo, Antonio, and Bhagwan Chowdhry, 2002, Resources, real options, and corporate strate
Journal of Financial Economics 63, 211-234.
Black, Fischer, 1976, Studies of stock price volatility changes, in proceedings of the Business
Economic Statistics Section, American Statistics Association.
Bollen, Nicolas, 1999, Real options and product life cycles, Management Science 45, 670-684.
Brandt, Michael, Runeet Kishore, Pedro Santa-Clara, and Mohan Venkatachalam, 2009, Earn
announcements are full of surprises, Working paper, Duke University.
Brennan, Michael, and Eduardo Schwartz, 1985, Evaluating natural resource investments, J
nal of Business 58, 135-157.
Brown, Gregory, and Nishad Kapadia, 2007, Firm-specific risk and equity market developm
Bulan, Laarni, 2005, Real options, irreversible investment and firm uncertainty: New evid
from U.S. firms, Review of Financial Economics 14, 255-279.
Caballero, Ricardo, and Robert Pindyck, 1992, Uncertainty, investment, and industry evolut
Working paper, MIT.
Campbell, John, and Ludger Hentschel, 1992, No news is good news: An asymmetric mode
changing volatility in stock returns, Journal of Financial Economics 31, 281-318.
Campbell, John, Martin Lettau, Burton Malkiel, and Yexiao Xu, 2001, Have individual st
become more volatile? An empirical exploration of idiosyncratic risk, Journal of Finance
1-44.
Cao, Charles, Timothy Simin, and Jing Zhao, 2008, Can growth options e
idiosyncratic risk? Review of Financial Studies 21, 2599-2633.
Carlson, Murray, Adlai Fisher, and Ron Giammarino, 2004, Corporate inve
dynamics: Implications for the cross-section of returns, Journal of Finan
Carlson, Murray, Adlai Fisher, and Ron Giammarino, 2006, Corporate inve
dynamics: Implications for SEO event studies and long-run performance, J
61, 1009-1034.
Carlson, Murray, Adlai Fisher, and Ron Giammarino, 2010, SEO risk dynamics, Review of Finan
cial Studies 23, 4026-^1077.
Carlson, Murray, Zeigham Khokher, and Sheridan Titman, 2007, Equilibrium exhaustible resource
price dynamics, Journal of Finance 62, 1663-1703.
Chemmanur, Thomas, Shan He, and Debarshi Nandy, 2010, The going public decision and the
product market, Review of Financial Studies 23, 1855-1908.
Chen, Jason, Marcin Kasperczyk, and Hernan Ortiz-Molina, 2011, Labor unions, operating flexi
bility, and the cost of equity, Journal of Financial and Quantitative Analysis 46, 25-58.
Christie, Andrew, 1982, The stochastic behavior of common stock variances: Value, leverage and
interest rate effects, Journal of Financial Economics 10, 407^132.
Cooper, Michael, Huseyin Gulen, and Michael Schill, 2008, Asset growth and the cross-section of
stock returns, Journal of Finance 63, 1609-1652.
Da, Zhi, Re Guo, and Ravi Jagannathan, 2012, CAPM for estimating the cost of equity capital:
Interpreting the empirical evidence, Journal of Financial Economics 103, 204-220.
Duffee, Gregory, 1995, Stock return and volatility. A firm-level analysis, Journal of Financial
Economics 37, 399-420.
Duffee, Gregory, 2002, Balance sheet explanations for asymmetric volatility, Working paper, Uni
versity of California.
Fama, Eugene, and Kenneth French, 1993, Common risk factors in the returns of stocks and bonds,
Fama, Eugene, and Kenneth French, 1997, Industry costs of equity, Journal of Financial Economics
49, 153-193.
Fama, Eugene, and James MacBeth, 1973, Risk, return, and equilibrium: Empirical tests, Journal
of Political Economy 81, 607-636.
Fan, Ying, and Lei Zhu, 2010, A real options-based model and its application to China's overseas
oil investment decisions, Energy Economics 32, 627-637.
Frazzini, Andrea, and Owen Lamont, 2007, The earnings announcement premium and trading
volume, Working paper, University of Chicago.
French, Kenneth, William Schwert, and Robert Stambaugh, 1987, Expected stock returns and
volatility, Journal of Financial Economics 19, 3-29.
Galai, Dan, and Ronald Masulis, 1976, The option pricing model and the risk factor of stock,
Ghosal, Vivek, 1991, Demand uncertainty and the capital-labor ratio: Evidence from the U.S.
manufacturing sector, Review of Economics and Statistics 73, 157-161.
Gibbons, Michael, Stephen Ross, and Jay Shanken, 1989, The test of the efficiency of a given
portfolio, Econometrica 57, 1121-1152.
Gramm, Cynthia, and John Schnell, 2001, The use of flexible staffing arrangements in core pro
duction jobs, Industrial and Labor Relations Review 54, 245-268.
Griffin, John, Jeffrey Harris, Tao Shu, and Selim Topaloglu, 2011, Who drove and burst the tech
bubble? Journal of Finance 66, 1251-1290.
Guiso, Luigi, and Giuseppe Parigi, 1999, Investment and demand uncertainty, Quarterly Journal
of Economics 114, 185-227.
Hirsh, Barry, and David Macpherson, 2003, Union membership and coverage database from the
current population survey: Note, Industrial and Labor Relations Review 56, 349-354.
Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to buying winners and selling losers:
Implications for stock market efficiency, Journal of Finance 48, 65-91.
Joos, Philip, and Alexei Zhdanov, 2008, Earnings and equity valuation in the biotech industry:
Theory and evidence, Financial Management 37, 431-459.
Kapadia, Nishad, 2007, The next Microsoft? Skewness, idiosyncratic volatility, and expecte
turns, Working paper, Rice University.
Karpoff, Jonathan, 1987, The relation between price changes and trading volume: A survey, Jo
nal of Financial and Quantitative Analysis 22, 109-126.
Leahy, John, and Toni Whited, 1996, The effect of uncertainty on investment: Some stylized fa
Journal of Money, Credit, and Banking 28, 6483.
Lemmon, Michael, and Jaime Zender, 2010, Debt capacity and tests of capital structure theo
Journal of Financial and Quantitative Analysis 45, 1161-1187.
Loughran, Tim, and Jay Ritter, 1995, The new issues puzzle, Journal of Finance 50, 23-51.
Lyandres, Evgeny, Le Sun, and Lu Zhang, 2008, The new issues puzzle: Testing the investm
based explanation, Review of Financial Studies 21, 2825-2855.
Lyon, John, Brad Barber, and Chih-Ling Tsai, 1999, Improved methods for tests of long-run a
mal stock returns, Journal of Finance 54, 165-201.
Majd, Saman, and Robert Pindyck, 1987, Time to build, option value, and investment decisio
Marschak, Thomas, and Richard Nelson, 1962, Flexibility, uncertainty, and economic theory
ternational Review of Economics 14, 42-58.
McDonald, Robert, and Daniel Siegel, 1985, Investment and valuation of firms when there is
option to shut down, International Economic Review 26, 331-349.
McDonald, Robert, and Daniel Siegel, 1986, The value of waiting to invest, Quarterly Journa
Economics 101, 707-727.
Mills, David, 1984, Demand fluctuations and endogenous firm flexibility, Journal of Indus
Economics 33, 55-71.
Moel, Alberto, and Peter Tufano, 2002, When are real options exercised? An empirical stud
mine closings, Review of Financial Studies 15, 35-64.
Newey, Whitney, and Kenneth West, 1987, A simple, positive semi-definite, heteroskedasticity
autocorrelation consistent covariance matrix, Econometrica 55, 703-708.
Oi, Walter, 1961, The desirability of price instability under perfect competition, Econometrica
58-64.
Ottoo, Richard, 1998, Valuation of internal growth opportunities: The case of a biotechnology
company, Quarterly Review of Economics and Finance 38, 615-633.
Paddock, James, Daniel Siegel, and James Smith, 1988, Option valuation of claims on real assets:
The case of offshore petroleum leases, Quarterly Journal of Economics 103, 479-508.
Petersen, Mitchell, 2009, Estimating standard errors in finance panel data sets: Comparing ap
proaches, Review of Financial Studies 22, 435-480.
Pindyck, Robert, 1988, Irreversible investment, capacity choice, and the value of the firm, American
Economic Review 78, 969-985.
Pindyck, Robert, 1993, A note on competitive investment under uncertainty, American Economic
Review 83, 273-277.
Ritter, Jay, 2003, Investment banking and security issuance, in George Constantinides, Mil
ton Harris, and Rene Stulz, eds.: Handbook of Economics and Finance (North Holland,
Amsterdam).
Scheinkman, Jose, and Wei Xiong, 2003, Overconfidence and speculative bubbles, Journal of Po
litical Economy 111, 1183-1219.
Schultz, Paul, 2008, Downward-sloping demand curves, the supply of shares, and the collapse of
internet stock prices, Journal of Finance 63, 351-378.
Stigler, George, 1939, Production and distribution in the short run, Journal of Political Economy
47, 305-327.
Veronesi, Pietro, 1999, Stock market overreactions to bad news in good times: A rational expecta
tions equilibrium model, Review of Financial Studies 12, 975-1007.
Whited, Toni, 2006, External finance constraints and the intertemporal pattern of intermittent
investment, Journal of Financial Economics 81, 467-502.
Xing, Yuhang, 2008, Interpreting the value effect through the Q-theory: An empirical investigation,
Review of Financial Studies 21, 1767-1795.

23261366

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

23261366

Загружено:

Авторское право:

Доступные форматы

American Finance Association

Real Options, Volatility, and Stock Returns

Real Options, Volatility, and Stock Returns

GUSTAVO GRULLON, EVGENY LYANDRES, and ALEXEI ZHDANOV*

It is well established in the asset pricing literature that aggregat

Contrary to the evidence at the aggregate level, however, Du

whether proxies for operating flexibility can explain cross-sectional differen

CAPM. We exploit this result to test our main hypothesis. If

The remainder of the paper is organized as follows. In Section I we dis

I. Measure of Volatility, Data Sources, and Summary Statistics

estimated volatility in month t and the estimated volatility in

AVOLu = VOLlt - YOU,,_x. (2)

B. Main Data Sources and Summary Statistics

Mean SD P5 Median P95 # Obs.

Excess return 0.603 17.731 -21.709 -0.334 27.442 3,046,524

II. Return-Volatility Relation and Investment Opportunities

1"i,t ~ rf,t = <*t + Pt&VOLi t + YtVMKTi.t + &t %i,t + i,t- (3)

Returns and Contemporaneous Changes in Volatility

Market factor loading 0.298 0.217 0.300 0.2155 0.171 0.1075

credit if available, minus post-retirement benefit assets if available. If stock

t 7*f x ~t~ VMKTi t^Tm,r ^f.r) 4" (4)

This table shows that contemporaneous changes in firm-level volatility are p

and to firm value, we measure future sales growth starting fr

ri,t rf,t = ott + faAVOLij + vtGRijAVOLi t + YtVMKTi,, + &t%i,t + (5)

Returns, Contemporaneous Changes in Volatility, and Investment

Proxy for Investment Opportunities

Market 0.120 0.106 0.147 0.134

III. Return-Volatility Relation, Convexity, and Flexibility

where V (x) is firm value as a function of variable x having a standard deviation

E(7r(p)) = 7T(E(p)) + (7)

Abnj-i,r = oiij + PijEarrusurpi x + Yi,tEarnJiurpi x2 + eia, (8)

where Earnia is the firm's earnings per share in qua

ECEarnitX) = Earni:X-i + y^(Earni<r-n - Earnitr-n-4)/8. (10)

The measure of convexity of firm value to its earnings surprise is the es

Abnj-it = ctij + Pi,tSalesjsurpix + Yi,tSalesJsurpix2 + e;,T, (11)

where Salesjsurpi r is estimated similar to Earnjsurpl T in equation (9). The

Returns, Contemporaneous Changes in Volatility, and Measures

Proxy for Flexibility

Earnings Convexity Sales Convexity Union Membership

Market 0.103 0.112 0.161

Low flexibility 2.01% 1.89% 1.65%

of their value represented by real options and in a lower sensitivity of their

operating in industries with lower union membership are more sensitive to

IV. Evolution of the Return-Volatility Relation around Real Option

The Relation between Returns and Changes in Volatility around

Panel A: Investment spikes

Issuing firm 1.232 1.254 1.021 0.691 1.100

Years since SEO

Issuing firm 1.035 1.874 0.212 0.572

Panel C: Financing spikes

Issuing firm 0.724 1.253 0.711 0.583 1.270

Similar to the investment spikebased tests, we want to ensure that the

V. Industry Analysis of the Return-Volatility Relation

An additional way of examining whether the positive relation between firm

defer investment, expand, contract, shut down, and restart operations ac

We define Fama and French (1997) industries 27 (precious metals), 28 (min

B. Oil and Gas Industry

Returns, Contemporaneous Changes in Volatility, and Industry-Wide

of firm value to changes in volatility of a process underlying its real options.

options, we are able to construct more direct measures of the proportion of r

+ 7 xiit +i,t, (12)

8 We convert cubic feet of natural gas to a barr

Returns of Oil and Gas Firms and Contemporaneous Changes

Type of undeveloped reserves