Volatility Information Trading in The Option Market

THE JOURNAL OF FINANCE VOL. LXIII, NO.
3 JUNE 2008
Volatility Information Trading

in the Option Market
SOPHIE X. NI, JUN PAN, and ALLEN M. POTESHMAN
ABSTRACT
This paper investigates informed trading on stock volatility in the option market. We
construct non-market maker net demand for volatility from the trading volume of
individual equity options and find that this demand is informative about the future
realized volatility of underlying stocks. We also find that the impact of volatility demand on option prices is positive. More importantly, the price impact increases by 40%
as informational asymmetry about stock volatility intensifies in the days leading up
to earnings announcements and diminishes to its normal level soon after the volatility
uncertainty is resolved.
THE LAST SEVERAL DECADES have witnessed astonishing growth in the market for
derivatives. The markets current size of $200 trillion is more than 100 times
greater than 30 years ago (Stulz (2004)). Accompanying this impressive growth
in size has been an equally impressive growth in variety: The derivatives market now covers a broad spectrum of risk, including equity risk, interest rate risk,
weather risk, and, most recently, credit risk and inf lation risk. This phenomenal growth in size and breadth underscores the role of derivatives in financial
markets and their economic value. While financial theory has traditionally emphasized the spanning properties of derivatives and their consequent ability to
improve risk-sharing (Arrow (1964) and Ross (1976)), the role of derivatives as a
vehicle for the trading of informed investors has emerged as another important
economic function of these securities (Black (1975) and Grossman (1977)).
We contribute to the body of knowledge on the economic value of derivatives
by investigating the role of options as a mechanism for trading on information about future equity volatility. Our focus on informed volatility trading
is motivated to a large extent by the fact that equity options are uniquely
suited to investors with information about future volatility. Unlike traders with
Ni is at the Hong Kong University of Science and Technology. Pan is at the MIT Sloan School
of Management and NBER. Poteshman is at the University of Illinois at Urbana-Champaign. We
thank Joe Levin, Eileen Smith, and Dick Thaler for assistance with the data used in this paper.
We thank Bob Whaley and an anonymous referee for extensive and insightful comments. We also
benefited from the comments of Joe Chen, Jun Liu, Neil Pearson, Josh Pollet, Rob Stambaugh
(the editor), Dimitri Vayanos, Jiang Wang, Josh White, and seminar participants at the University
of British Columbia, the University of Illinois at Urbana-Champaign, the University of Virginia,
Vanderbilt University, and the 2006 AFA meetings. Poteshman thanks the Office for Futures and
Options Research at the University of Illinois at Urbana-Champaign for financial support. We bear
full responsibility for any remaining errors.
1059
1060
The Journal of Finance
directional information about underlying stock prices who can trade in either
the stock or option markets, traders with volatility information can only use
nonlinear securities such as options. Moreover, while the question of whether
traders use options to trade on directional information has been examined in
some detail (Stephan and Whaley (1990), Amin and Lee (1997), Easley, OHara,
and Srinivas (1998), Chan, Chung, and Fong (2002), Chakravarty, Gulen, and
Mayhew (2004), Cao, Chen, and Griffin (2005), and Pan and Poteshman (2006)),
the analogous question about volatility information trading has not been systematically addressed in the literature.1 Since volatility plays such a central
role in both the pricing of options and the reasons for trading options, a better
understanding of volatility information trading is clearly important.
Our empirical investigation takes advantage of a unique data set from the
Chicago Board Options Exchange (CBOE) that records purchases and sales of
put and call options by non-market makers over the 1990 to 2001 period. For
each underlying stock, we construct daily non-market maker net demand for
volatility. Motivated by theoretical models of private information trading, particularly those of Kyle (1985) and Backs (1993) extension of it to a setting that
include options, we pursue two strands of empirical investigation to examine
the presence of volatility information trading in the option market. First, we
investigate the extent to which the volatility demand extracted from the option
market predicts the future volatility realized by underlying stocks. Second, we
examine the price impact of volatility demand, focusing especially on the time
variation of the price impact leading up to earnings announcements when the
level of informational asymmetry is high. Both of these strands yield evidence
in support of volatility information trading in the option market.
Our first main finding is that option market demand for volatility predicts
the future realized volatility of underlying stocks even after controlling for
option implied volatility and a number of other variables. The predictability
lasts at least 1 week into the future and is robust to different measures of
realized volatility and controls for directional information in the option volume.
A natural interpretation of the evidence that option volume is informative about
future volatility is that investors trade on volatility information in the option
market and the information is subsequently ref lected in the underlying stocks.
1
The mass media, however, often attributes option market activity to volatility trading. For
example, on October 5, 2006, the Wall Street Journal reported heavy trading of McDonalds options (Hodi (2006f)): Nearly 100,000 calls and 94,000 put options changed hands, many of them
simultaneously, in a position called a straddle. A straddle involves buying a put and a call at the
same strike price, and profits from a large move in the stock in either direction. More interestingly, a large fraction of these straddles targeted the November contracts. The catalyst for the
trading? The Company is due to report September sales on Oct. 12. McDonalds tends to preview
quarterly earnings during such reports. The company is expected to report detailed earnings for
the third-quarter later this month (Hadi (2006f)). For further examples, see Hadi (2006a2006e).
Lakonishok, Lee, Pearson, and Poteshman (2007) provide evidence that only a small percentage of
the option trading of customers of a discount brokerge house could be part of volatility trades such
as straddles or strangles. Pearson, Poteshman, and White (2007) show that option trading impacts
the volatility of underlying stocks.
Volatility Information Trading in the Option Market
1061
This interpretation is corroborated by two additional results. First, we find

that the non-market maker net demand for volatility constructed from open option volume (trades in which non-market makers open brand new option positions) has stronger predictability in both magnitude and statistical significance
than that constructed from close option volume (trades in which non-market
makers close existing option positions). While both open and close trades can be
informationally motivated, the informational content of close trades is expected
to be lower because traders can only use information to close positions if they
happen to have appropriate positions open at the time they become informed.
The stronger predictability of the net demand constructed from open volume
is therefore consistent with the predictability we identify having its economic
source in volatility information trading in the option market.2 Second, we find
that the net demand for volatility constructed from option volume that could
have been part of straddle trades is a stronger predictor of future volatility
than net demand for volatility constructed from option volume that could not
have been part of straddle trades. Since straddles are the leading strategy for
trading on volatility, the stronger predictability of the volume with the higher
concentration of straddle trading is also consistent with the predictability we
identify originating from informed volatility trading in the option market.
Our second main finding is that volatility demand has a positive and timevarying price impact that increases on days leading up to earnings announcements. In standard information trading models such as that in Kyle (1985),
the presence of informational asymmetry induces the market maker to adjust
price in response to demand, resulting in a positive relation between equilibrium price and demand. Our empirical analysis shows that the net demand for
volatility indeed positively impacts option prices.
Although the finding of a positive price impact confirms a prediction of the
hypothesis that option investors trade on private volatility information, it could
also be produced by demand pressure that is unrelated to informational asymmetry. We offer two additional results to assess whether the positive price impact is produced by demand pressure alone. In both analyses, we take advantage of the fact that unlike pure demand pressure, the price impact caused by
informational asymmetry depends in an important way on the level of informational asymmetry. We first employ earnings announcements to exploit time
variation in informational asymmetry. Prescheduled earnings announcements
are exogenous information events. Thus, in the period leading up to earnings
announcements days (EADs), option market makers face increasing informational asymmetry. While there will be informational asymmetry about both the
directional move and the volatility of the underlying stock price at the EAD,
the asymmetry about volatility information is likely to be of more concern to
2
An alternative explanation for this result is that closing trades are more likely to be in-themoney or out-of-the-money, both of which are less sensitive to volatility. To check whether this is the
case, we compared the relative predictability within each category of moneyness and found that the
predictability from opening option transactions dominates that from closing option transactions
regardless of the moneyness category.
1062
option market makers.3 Our results show that the price impact per unit of nonmarket maker net demand for volatility increases as informational asymmetry
increases in the week leading up to EADs. By the day before EADs, the price
impact per unit of volatility demand increases to 40% above its normal level.
Following EADs, the additional price impact diminishes and becomes statistically insignificant. Since the time variation in the price impact mirrors the
time variation in informational asymmetry, this evidence is consistent with
market makers protecting themselves from option market trades motivated by
volatility information.4
To further address the alternative hypothesis of demand pressure, we make
use of the fact that volume that opens and closes option positions is associated
with different levels of informational asymmetry. For example, in our predictive analysis, we find that the non-market maker net demand for volatility
constructed from open option volume has stronger predictability than that
from close option volume, indicating a higher informational content for the
open volume. Similarly, Pan and Poteshman (2006) show that open volume
contains more information than close volume about the future direction of underlying stock prices. If the market makers possess any ability to distinguish
open trades from close trades, they will react more to open volume because of
its greater informational content.5 This additional price impact is unrelated
to pure demand pressure and can be attributed to informational asymmetry. Our analysis indicates that open volume does indeed have a greater price
impact.
Our paper relates to several strands of the empirical option literature. First,
our predictive analysis is related to the literature that examines the informational content of option-implied volatility for future realized volatility.6 A
common finding in these papers is that the implied volatility subsumes other
publicly available information, including the past-realized volatility (e.g., see
Christensen and Prabhala (1998)). The reason that implied volatility provides
better predictions may be that information about future volatility is impounded
3
The market makers can delta-hedge (with the underlying stock) the risk associated with the
directional move, but collectively they must bear the volatility risk. Moreover, informed traders
with directional information can trade in either the option or the stock market, while traders with
volatility information can only trade in the option market.
4
It is also interesting to note that the predictive power of net volatility demand approximately
doubles on the day before EADs compared to the average level of 1-day ahead predictability. This
result is consistent with the hypothesis that volatility demand contains more information before
EADs.
5
Although the market makers do not directly observe in real time whether trades open or close
option positions, at the end of the day they can infer the breakdown of open versus close trades
during the day from changes in open interest. It seems likely that they would develop some realtime ability to assess the probability that trades coming into the market are opening or closing
positions.
6
See, for example, Latane and Rendleman (1976), Canina and Figlewski (1993), Christensen and
Prabhala (1998), Day and Lewis (1992), Jorion (1995), Lamoureux and Lastrapes (1993), Poteshman (2000), Ederington and Guan (2002), and Chernov (2002). In a similar vein, Ederington and
Lee (1996) investigate the impact of news on option implied volatility.
1063
into option prices through the trading process. By using quantity information in
addition to price information, our predictive analysis complements this literature by examining the process of information incorporation at the level of option
trading. In particular, we find that even though the non-market maker net demand for volatility is observable by market makers, it provides information
about future volatility beyond what is contained in option-implied volatility.7
Our findings are also related to a number of studies that examine stock option
prices and trading volume in the time surrounding earnings announcements
(Patell and Wolfson (1979, 1981) and Whaley and Cheung (1982)). For example, Amin and Lee (1997) investigate abnormal trading volume in the option
market around the announcement of earnings news and provide evidence of
directional information trading in the option market. Donders, Kouwenberg,
and Vorst (2000) and Dubinsky and Johannes (2005) document the impact of
earnings announcements on equity option prices and find that implied volatility increases before EADs and drops afterward, as suggested in Ederington
and Lee (1996) for prescheduled events. Dubinsky and Johannes (2005) also
develop an option pricing model to incorporate jumps on EADs and find that
firms with high option-implied estimates of earnings uncertainty subsequently
have volatile stock returns on EADs. This long line of research attests to the
importance of earnings announcements as informational events in the equity
option market. Our analysis adds to this body of literature by addressing time
variation in the level of informational asymmetry surrounding these announcements. In particular, we provide empirical evidence to link time variation in
informational asymmetry directly to time variation in the price impact in the
equity option market.
Finally, our findings on the option price impact of volatility demand are related to two recent papers that investigate the relationship between the demand for options and the prices of options. Bollen and Whaley (2004) show that
option-implied volatilities are positively related to the net buying pressure for
options, consistent with our finding of a positive price impact. They also find
that approximately 20% of the price impact is reversed the next day, suggesting
that price pressure is one component of the positive price impact. By contrast,
the main objective of the second strand of our empirical work is to determine
whether there is a component of price impact that relates to volatility information trading. We do so by investigating the connection between time variation
in price impact and time variation in informational asymmetry as well as the
differential price impact of open and close option volume. Our empirical results
are also related to the findings of Garleanu,

Pedersen, and Poteshman (2006),
who develop a theoretical model linking the level of aggregate demand to the
level of option prices and provide empirical support for their models predic
tions. The Garleanu
et al. (2006) paper is agnostic about the source of option
demand. We provide evidence that the demand comes in part from investors
7
A caveat here is that the papers in this literature typically predict the realized volatility over
the life of the option from which volatility is implied while we predict volatility over only one future
trading day.
1064
with volatility information and that this component of demand impacts option
prices.
The rest of the paper is organized as follows. Section I develops our empirical
specifications. Section II details the data, and Section III presents the results.
Finally, Section IV concludes.
I. Empirical Specification
A. Information in Option Volume for Future Stock Volatility
The first part of our empirical investigation examines whether option volume possesses information about the future volatility of underlying stocks. If
informed investors indeed bring private information about future volatility to
the option market, then one would expect the non-market maker net demand
for volatility to be positively related to the future volatility of underlying stocks.
We construct the demand for volatility from the market makers perspective
and separate trading volume into non-market maker buys and sells of call
and put options. Both call and put options have positive vega (exposure to
volatility), so we treat buy volume for both call and put options as positive
demand for volatility and sell volume as negative demand for volatility. Since
options of varying strike prices and maturities have differing sensitivities to
changes in volatility, we weight the demand for each contract by the return to
the option per unit change in volatility. Specifically, for each stock i on day t,
the daily demand for volatility is measured by
Di,t
K ,T
ln Ci,t
K
i,t
K ,T
K ,T
BuyCalli,t
SellCalli,t
K ,T
ln Pi,t
K
i,t

K ,T
K ,T
BuyPuti,t
,
SellPuti,t
(1)
where CK,T
i,t is the price at time t of the call on underlying stock i with strike price
K and maturity T; PK,T
i,t is the price for the similar put; i,t is the volatility of underlying stock i at time t; BuyCallK,T
i,t is the number of call contracts purchased
by non-market makers on day t on underlying stock i with strike price K and
K,T
K,T
maturity T; and SellCallK,T
i,t , BuyPuti,t , and SellPuti,t are the analogous quantities for, respectively, the sale of calls and the purchase and sale of puts. The
summations in equation (1) are over all strikes and all maturities greater than
1 week that are available for underlying stock i on trade day t.8 In the empiriK ,T
K ,T
K ,T
cal work, we approximate ln Ci,t
/i,t with (1/Ci,t
)BlackScholesCallVegai,t
K ,T
(and similarly for ln Pi,t
/i,t ), where the Black-Scholes vega is computed
with the volatility of the underlying stock set to the sample volatility from the
We follow the common practice of excluding options with very short maturities. All of our results
are robust to including these options.
1065
60 trading days of returns leading up to t. It should be noted that our measure

of demand for volatility is observable by market makers who can see the buy
and sell orders coming into the market, but it is not observable by other option
market participants. Although a non-market maker could use quotation and
transaction data in conjunction with the Lee and Ready (1991) algorithm to
classify option volume as buyer- or seller-initiated, the classification is only on
the order of 80% accurate (Savickas and Wilson (2003)).
We test whether the non-market maker net demand for volatility in the option
market predicts the future volatility of underlying stocks by estimating the
following specification:
OneDayRV i,t = a + b Di,t

j + c Di,t j Indi,t
+ d OneDayRV i,t1 + e OneDayRV i,t1 Indi,t

+ f OneDayRV i,t2 + g OneDayRV i,t2 Indi,t
+ h OneDayRV i,t3 + i OneDayRV i,t3 Indi,t
+ j OneDayRV i,t4 + k OneDayRV i,t4 Indi,t
+ l OneDayRV i,t5 + m OneDayRV i,t5 Indi,t

+ n Indi,t + o I Vi,t1 + p I Vi,t1 Indi,t + q abs Di,t
j

+ r abs Di,t
+
s
optVolume
Ind
i,t
i,t j
j
+ t optVolumei,t j Indi,t + u ln(stkVolumei,t j )
+ v ln(stkVolumei,t j ) Indi,t + i,t ,
(2)
where OneDayRV i,t is a proxy for the realized volatility of stock i on date t,
Indi,t is one if t is an EAD for stock i and zero otherwise, IV i,t is the average
implied volatility of the shortest maturity (of at least 5 calendar days) closest to at-the-money (ATM) call and put on underlying stock i on trade day t,
D
i,t is the delta-weighted sum of all of the non-market maker option trading
volume on underlying stock i on trade day t, optVolumei,t is the number of option contracts that trade on underlying stock i on day t, and stkVolumei,t is the
number of shares of stock i that trade on day t. The regression specified in equation (2) is estimated separately for different values of j, and for a fixed j uses
the net demand for volatility at time t j to predict the j-day-ahead realized
stock volatility. According to the hypothesis that volatility information trading
exists in the option market, we would expect the coefficient estimate on Di,tj
to be positive and significant for at least some of the j-day-ahead predictive regressions. Moreover, assuming that on average informed traders possess more
information about what will occur at EADs than on non-EADs, we would expect
to see an incremental predictability from demand for volatility, that is, a positive slope coefficient on the interaction term Di,tj Indi,t in equation (2). The
Indi,t term that appears by itself on the right-hand side of equation (2) controls
for differences in 1-day realized volatility on EADs that are unrelated to volatility demand. The terms involving OneDayRV i,tn for n = 1, 2, . . . , 5 control for
1066
GARCH type volatility clustering. The IV i,t1 terms control for publicly available information about future realized volatility that should all be impounded
into the option prices, and the terms involving optVolumei,tj and stkVolumei,tj
control for any relationship between option or stock volume and future volatility
and also for the possibility that option or stock volume is correlated with option
net demand for volatility. The empirical work below will also include specifications that add underlying stock characteristics and interactions of underlying
stock characteristics with the Di,tj variable in order to investigate potential
cross-sectional heterogeneity in volatility information trading.
Clearly, investors trade options for reasons other than possessing private
information about the volatility of underlying stocks (e.g., for hedging or for
liquidity reasons). In general, this fact should just make it more difficult to
detect a relationship between option market demand for volatility and the future volatility of underlying stocks. It does, however, seem important to control
for investors trading in the option market on private information that the underlying stock price will increase or decrease. Pan and Poteshman (2006) show
that option volume contains information about the future direction of underlying stock price movements. As a result, it is possible that a positive relation
between non-market maker net demand for volatility and future stock price
volatility could be driven by investors trading on directional information in
the option market. This would occur, for example, if investors with positive directional information about an underlying stock buy calls and the stock price
subsequently increases. The call purchases would increase the net demand variable while the later increase in stock price would increase the measure of future
realized volatility. For negative information, the purchase of puts would also
increase the net demand variable and the later decrease in stock price would
increase the measure of future realized volatility as well.
We control for option market trading on directional information through the
terms in specification (2) involving abs(D
i,tj ), the delta-weighted sum of all of
the non-market maker option trading volume. When computing this variable,
buy volume gets a positive sign, sell volume gets a negative sign, and the delta
(positive for calls, negative for puts) is computed under the Black-Scholes assumptions with the volatility of the underlying stock set to the sample volatility
from the 60 trading days of returns leading up to t. Hence, D
i,t measures the net
equivalent number of shares purchased by non-market makers through their
option market trading on underlying stock i on trade date t. Our control uses
the absolute value of this quantity, because larger absolute positive (negative)
values are consistent with larger quantities of positive (negative) directional
trading in the option market.
Aside from option trading that is based on directional information, it is not obvious how option activity that is not motivated by volatility information would
bias our tests toward spurious findings of informed volatility trading. For example, option trading motivated by hedging or liquidity concerns should just
add noise to our tests, making it more difficult to detect any effects. Nonetheless, we also run our tests after constructing the demand for volatility variable
only from option volume that could have been part of straddle trades or that
1067
could not have been part of straddle trades. Since volume that is potentially
part of straddle trades will have a higher concentration of volatility-based trading than volume that could not have been part of straddle trades, the former
should give stronger results than the latter if, indeed, option investors trade on
private volatility information.
B. Price Impact and Informational Asymmetry
The second part of our empirical investigation examines the response of option market makers to volatility demand. If market makers believe that option
order f low comes in part from investors with private volatility information,
then they will increase (decrease) option prices in response to positive (negative) volatility demand.
In order to test for these option pricing implications of informed trading on
volatility information, we construct securities out of options that have high
sensitivity to realizations of equity volatility and low sensitivity to directional
moves in the underlying stock. We do so by forming near-the-money straddles.
On each day t and for each stock i, we pick a pair of call and put options with the
same time to expiration and the same strike price that are as close to the money
as possible. This yields close to a clean security on volatility: While both call
and put options increase in value with increasing volatility, their respective
sensitivities to directional moves by the underlying stock are opposite in sign
and approximately equal in magnitude if both options are ATM. However, given
that we are using exchange-traded options, whose available moneyness and
time-to-expiration vary over time, we measure the straddle price in terms of
Black-Scholes implied volatility. Specifically, we convert the market-observed
p
call and put option prices into their respective implied volatility, IV ci,t and IV i,t , 9
and measure the price of the security by

1 c
p
IV i,t =
(3)
IV i,t + IV i,t .
2
The empirical specification used to investigate the impact of volatility demand is as follows:

IV i,t IV i,t1 IV i,t1

EAD
EAD
= + Di,t
+ Di,t
Ind EAD 0 i,t
5 Ind EAD 5 i,t + + 0

EAD
+ + EAD
Ind EAD 5 i,t +
+ 5
+5 Ind EAD + 5 i,t

EAD
+ 0EAD Ind EAD 0 i,t + + +5
Ind EAD + 5 i,t + i,t ,
(4)
where Di,t is the non-market maker net demand for volatility on underlying
stock i on trade date t. The corresponding slope coefficient, , in equation (4)
therefore captures the average price impact of a one-unit increase in volatility
9
Specifically, we use binomial tree option pricing, taking into account dividend payments and
early exercise premiums. See, for example, Harvey and Whaley (1992) for details on extracting
volatility information using the dividend-adjusted binomial method on American-style options.
1068
demand. The economic interpretation of this price impact variable could be

twofold. While market makers may adjust option prices to protect themselves
from informational asymmetry, even in the absence of informational asymmetry
they may require a premium (discount) to absorb positive (negative) order f low.
As a result, the price impact variable may potentially contain two components:
one corresponding to informational asymmetry and the other to pure demand
pressure.
To determine whether any of the price impact comes from informational
asymmetry, we take advantage of the fact that the noninformational price
impact is insensitive to the level of informational asymmetry, while the
information-based component depends on the level of informational asymmetry
in a crucial way. We exploit this fact in two ways to assess whether any price
impact is attributable to volatility information.
First, in addition to the demand for volatility constructed from all option
,Open
volume, we include another measure of volatility demand, Di,t
, which is
constructed only from the option volume attributable to non-market makers
opening new positions. One would naturally expect that an investor who acquires private information would be more likely to trade on it in the option
market by opening rather than closing positions.10 Hence, the new volatility
demand constructed from open volume is expected to have a greater concentration of informed traders. Provided that the market makers have some ability
to distinguish the open from the close transactions, the open demand with its
higher level of informational asymmetry is predicted to have a higher price
impact. Of course, if demand pressure alone is driving the price impact, then
the open volume demand variable will not exhibit any additional price impact.
,Open
The coefficient on Di,t
therefore provides a test to determine whether informational asymmetry contributes to the price impact.11
Second, we take advantage of the fact that the days leading up to earnings
announcements constitute natural exogenous events for which the level of informational asymmetry is likely to be increasing.12 If the information-based
explanation is an important component of the price impact, then we would observe a pattern of increasing price impact leading up to EADs. On the other
hand, if demand pressure is the dominant explanation, then we should not see
10
Pan and Poteshman (2006) show that this is indeed the case for directional information about
underlying stocks.
11
Market makers do not directly observe in real time whether orders coming into the market
open or close positions. However, they do observe the change in open interest at the end of every
day, and therefore presumably gain some ability to predict in real time whether order f low opens
or closes positions. To the extent that they do not develop this ability, it will be less likely that
volatility demand from open volume will have a significant coefficient estimate.
12
While there may be informational asymmetry about either the directional move of the underlying stock price or its volatility, the informational asymmetry with respect to volatility is likely to
be of more concern to option market makers. In contrast to the risk associated with the directional
move, which can be delta-hedged using the underlying stock, the volatility shock is a risk that
in aggregate the market makers have to bear. Moreover, informed traders with directional information can trade in either the option or the stock market, while informed traders with volatility
information can trade only in the option market.
1069
systematic variation in price impact around EADs. Equation (4) tests for this
variation by including Ind(EAD n)i,t , which is equal to one if trading day t + n
is an EAD for stock i and zero otherwise, and its interaction with the demand
variable. For example, Ind(EAD 5)i,t is one if t is 5 trade days before an EAD
for stock i and zero otherwise. The slope coefficient of the interaction term, EAD
for = 5, . . . , 1, 0, 1, . . . , 5, therefore captures the incremental price impact

of volatility demand around EADs. In particular, if informational asymmetry
contributes to the price impact, EAD
will be positive and significant on days
leading up to the EAD, but not for the days after the EAD when information
has already been released.
Finally, we add control variables in equation (4) including the (uninteracted)
Ind(EAD n)i,t variables to control for any systematic daily changes in IV
around EADs that are unrelated to volatility demand. We also include underlying stock characteristics as well as interactions of underlying stock characteristics with the Di,t variable in order to investigate potential cross-sectional
heterogeneity in the price impact of volatility demand.
II. Data
This section describes the data sources, sets out the criteria used in selecting
observations for the regressions specified in the previous section, and provides
summary statistics for the variables that appear in the regressions.
A. Data Sources
The data in this paper are drawn from a number of sources. The data used
to compute non-market maker net demand for volatility are obtained directly
from the CBOE. This data set contains daily non-market maker volume for all
CBOE-listed options over the period January 2, 1990 through December 31,
2001. For each option, the daily trading volume is subdivided into four types of
trades: open-buys, in which non-market makers buy options to open new long
positions, open-sells, in which non-market makers sell options to open new
written option positions, close-buys, in which non-market makers buy options
to close out existing written option positions, and close-sells, in which nonmarket makers sell options to close out existing long option positions. When
calculating the demand variable defined by equation (1), the buy volume is
composed of the open-buy and close-buy volume and the sell volume is composed of the open-sell and close-sell volume. The non-market maker volume
is also subdivided into four classes of investors: firm proprietary traders, public customers of full service brokers, public customers of discount brokers, and
other public customers. We use this subdivision when separating volume into
the volume that could have been part of straddle trades and the volume that
could not have been part of a straddle trade.
Over the period January 2, 1990 through January 31, 1996 we get option price
and volume data from the Berkeley Options Database, and from February 1,
1996 through December 31, 2001 we get price, volume, and implied volatility
1070
data from OptionMetrics. For the first part of the data period (i.e., through January 1996), we compute daily option implied volatilities from the midpoint of
the last bid-ask price quote before 3:00 p.m. Central Standard Time. In particular, we use the dividend-adjusted binomial method with the 1-month LIBOR
rate as a proxy for the risk-free rate and the actual dividends paid over the life
of an option as a proxy for the expected dividends. Starting in February 1996,
we use the implied volatilities supplied by OptionMetrics, which are computed
in a similar way.13
Stock prices, returns, volumes, and dividends during the entire sample period
are obtained from the Center for Research in Security Prices (CRSP). Earnings
announcement dates and the data for computing book equity are obtained from
Compustat.
B. Observation Selection
In order for there to be an observation when estimating regression specification (2) for an underlying stock i on trade date t, there must be data available to
construct non-market maker net demand for volatility on stock i on trade date
t j, and there must be data available to compute the option-implied volatility
on trade date t 1. In order for there to be an observation when estimating regression specification (4) for an underlying stock i on a trade date t, there must
be data available to construct non-market maker net demand for volatility on
stock i on trade date t, and there must be data available to construct the daily
change in implied volatility for underlying stock i from trade date t 1 to trade
date t.
Non-market maker net demand for volatility is considered available for underlying stock i on day t if there are at least 50 contracts of buy and sell trading
volume by non-market maker investors for options on underlying stock i on day
t.14 The data to construct the daily implied volatility for stock i on trade date t
are available if there exist a call and a put on stock i on trade date t that have
the same strike price and time to expiration that meet all of the following three
conditions: (1) both the call and put have strictly positive trading volume15 and
implied volatility data available on day t, (2) the time to expiration of the call
and the put is between 5 and 50 trading days, and (3) the ratio of the strike
price to the closing stock price on day t is between 0.80 and 1.20. If there is more
than one put-call pair satisfying the above three conditions, we choose the pair
whose strike price is nearest to the closing stock price on day t. If there is still
more than one put-call pair, the pair with the shortest maturity is selected. After selecting the put-call pair, we calculate the implied volatility by averaging
the call and put implied volatilities on day t. Data to construct the change in
13
However, when calculating implied volatilities OptionMetrics projects dividends rather than
using actual realized dividends and interpolates the risk-free rate from a zero curve constructed
from LIBOR rates and the settlement prices of CME Eurodollar futures.
14
The results reported below are similar if the cut off is set to 20 or 100 contracts.
15
We use the Berkeley Option Database to determine whether an option traded through January
1996. Beginning in February 1996 we use OptionMetrics to determine whether an option traded.
1071
implied volatility from day t 1 to day t is available if a put and a call that
meet the preceding criteria are available on both day t and t 1. In this case,
the daily change in implied volatility is computed by subtracting the average
call and put implied volatilities on day t 1 from the average on day t.
For specification (2), we normalize each stocks daily non-market maker net
demand for volatility, delta-adjusted net option volume, option volume, oneday realized volatility, and stock volume variables by subtracting the variables
calendar-year mean and then dividing by its calendar-year standard deviation.
When one of these variables has fewer than 25 time-series observations for
a stock in a calendar year, the stock is deleted for that year. For specification
(4), we normalize each stocks daily non-market maker net demand for volatility
variable in the same way as for specification (2), and we also de-mean the dependent (i.e., change in implied volatility divided by the level of implied volatility)
variable for each underlying stock i by subtracting its calendar-year mean.
Altogether, there are 703,229 stock-days and 2,220 different stocks over our
12-year period that meet the criteria for inclusion when estimating specification (2). There are 12,332 stock-days that fall on earnings announcement dates
(EADs) and 1,579 different stocks have observations on at least one EAD. There
are an average of 243 stocks on each trading day. We also rank selected stocks
into small, medium, and large size terciles on each trading day. The median market capitalization for the small tercile stocks falls into the fifth decile of NYSE
stocks, which ref lects the fact that stocks with active option trading tend to be
larger than average. The median market capitalization for our medium (large)
size tercile stocks fall into the ninth (tenth) size decile of NYSE stocks.
C. Summary Statistics
Table I contains summary statistics for the variables used in our main specifications. All statistics in this table are reported before any normalization but
after filtering for inclusion in specification (2). The average one-day realized
volatility (oneDayRV) is 551 basis points or 5.51%. The average implied volatility (IV) for all observations is 6,532 basis points or 65.32%. The smallest IV is
279 basis points or 2.79% and the largest is 49,999 basis points or 500%. IV
exhibits strong positive autocorrelation and positive skewness. We also report
the summary statistics for the daily change in IV (dIV) and the daily change
in IV divided by the level of implied volatility.
The vega weighted non-market maker net demand for volatility (D ) has a
mean value of 19.25. The fact that this statistic has an average value that is
negative implies that on average market makers are long volatility. However,
the much larger standard deviation and absolute values of the minimum and
maximum of D suggest that it is reasonable to think of the average volatility
demand as close to zero. The breakdown of the volatility demand into opening
and closing option volume also indicates that non-market makers on average
neither demand nor supply a large quantity of volatility when they open or
close option positions. In some of the tests below, we compute the volatility
according to whether the option volume could or could not have been part of
Table I
OneDayRV (bps)
Implied volatility (IV) (bps)
IV change (dIV) (bps)
dIV/IV (bps)
Volatility demand (D )
Open volatility demand
Close volatility demand
Straddle volatility demand
Nonstraddle volatility demand
Abs(D )
optVolume (contracts)
ln(stkVolume) (shares)
551
6,532
3.68
27.40
19.25
84.45
104
52.87
20.13
166
1,810
13.68
Mean
452
3,094
476
708
2,898
2,931
2,244
1,045
1,997
445
6,967
1.27
Std.
0.30
0.83
0.02
0.02
0.08
0.11
0.08
0.06
0.06
0.13
0.29
0.54
Auto
3.04
1.04
2.45
5.76
23.26
16.60
11.26
40.17
1.27
16.56
19.70
0.15
Skew
25.17
4.59
155
156
5,524
2,638
3,735
8,039
612
654
695
3.22
Kurt
0.00
279
12,796
7,633
299,868
334,503
438,727
43,641
149,116
0.00
50.00
6.91
Min.
14,185
49,999
33,945
33,777
703,889
560,901
289,835
222,005
128,863
45,197
503,480
19.58
Max.
This table reports summary statistics from the beginning of 1990 through the end of 2001 for the variables used in the papers tests. OneDayRV is
10,000 times the difference of an underlying stocks intraday high and low price divided by the closing stock price. Implied volatility (IV) is the average
implied volatilities of the selected straddles call and put. dIV is daily change in IV. Volatility Demand (D ) is the vega-weighted volatility demand
defined in equation (1), Open Volatility Demand is the vega-weighted volatility demand calculated from option volume that opens new positions, and
Close Volatility Demand is the vega-weighted volatility demand calculated from option volume that closes existing option positions. Straddle Volatility
Demand is the vega-weighted volatility demand from options that could have been part of straddle trades and Nonstraddle Volatility Demand is the
vega-weighted volatility demand from options that could not have been part of straddle trades. Abs(D ) is absolute net delta-weighted option volume,
optVolume is total number of option contracts traded, and ln(stkVolume) is log of daily stock volume.
Summary Statistics
1072
1073
straddle trades. Descriptive statistics for the demand variable computed in

these different ways are also provided in Table I.
Table II reports the average value of the main variables from Table I for
the quantiles of four underlying stock characteristics: relative option volume
(relOptVolume), historical stock volatility (histStockVol), the logarithm of market capitalization (ln(size)), and the book-to-market ratio (BM). Each of these
variables is computed once for each underlying stock for each calendar year.
The variable relOptVolume is 100 times the ratio of the years option volume
to stock volume, histStockVol is 100 times the sample standard deviation computed from the years daily stock returns, ln(size) is the natural logarithm of
year-end market capitalization, and BM is computed according to the method
of Fama and French (1992). In Table II and in the regressions reported below,
the value computed for relOptVolume, histStockVol, and ln(size) is associated
with the stock during the next calendar year. For BM, we follow Fama and
French (1992) and associate the value computed from data for a calendar year
with the stock starting the following June through the subsequent May. We
investigate the relOptVolume variable because of the evidence in Chakravarty
et al. (2004) that the ratio of option volume to stock volume impacts the degree of price discovery in the option market. We examine histStockVol because of the centrality to volatility in the option markets. The variables
ln(size) and BM are included, because of the widespread interest in these stock
characteristics.
The summary statistics in Table II show that the main volatility demand
variable does not display any clear pattern across the quantiles for any of the
four stock characteristic variables. The same is true for the volatility demand
variables constructed from open or close volume or from trades that could or
could not have been part of straddle trades.
Table III reports the cross-sectional correlation coefficients for the main variables in Table I. Implied volatility (IV) has close to zero correlation with volatility demand (D ). The daily change in implied volatility (dIV), on the other hand,
is positively correlated with volatility demand (0.22).
Figure 1 depicts the average values of the main variables leading up to, on,
and after EADs. In this figure, day 0 is an EAD, and date 10 (date +10) is
10 trading days before (after) an EAD. The horizontal line in each plot is the
average value of the variable across all observations. The first panel indicates
that, as expected, stocks have higher realized volatility on EADs. The second
panel reports the daily change in implied volatility divided by the level of implied volatility. (Since in the regression analysis below we use this quantity
measured in basis points, that is how we report it herefor example, if the
daily change in implied volatility is 1% of the level of implied volatility, the observation is recorded as 100 basis points.) Panel 2 indicates that before EADs,
daily changes in implied volatility are all positive. The largest average daily
change occurs one day before the EAD. On the EAD and for the following week
the daily changes in implied volatility are all negative. The day after the EAD is
the most negative with a value of about 300 basis points. After about a week,
the bars are not far from the sample mean.
1074
Panels 3 through 6 present information on the trading patterns in the stock

and option market around EADs. Interestingly, the volatility demand depicted
in Panel 3 is positive before the EAD and negative after the EAD. Hence, nonmarket makers are buying volatility before EADs and selling it after EADs.
Panels 5 and 6 indicate that increased activity both begins and ends sooner
Table II
Summary Statistics Broken Down by Stock Characteristics

This table reports the average values from the beginning of 1990 through the end of 2001 of a
number of variables for the quantiles of various stock characteristics. The variables are oneDayRV
(10,000 times the difference of an underlying stocks intraday high and low price divided by the
closing stock price), IV (the average implied volatilities of a selected straddles call and put), dIV
(the daily change in IV), dIV/IV (the daily change in the implied volatility divided by the level of implied volatility), D (the vega-weighted volatility demand defined in equation (1)), Open Volatility
Demand (the vega-weighted volatility demand calculated from option volume that opens new positions), Close Volatility Demand (the vega-weighted volatility demand calculated from option volume
that closes existing option positions), Straddle Volatility Demand (the vega-weighted volatility demand calculated from option volume that could have been part of straddles), and Nonstraddle
Volatility Demand (the vega-weighted volatility demand calculated from option volume that could
not have been part of straddles). The stock characteristics are: relOptVolume (100 times the ratio of
option volume to stock volume over the previous calendar year), histStockVol (100 times the sample
standard deviation of a stock over the past calendar year), ln(size) (the natural logarithm of the
market capitalization of the stock at the end of the previous calendar year), and BM (the stocks
book-to-market ratio as in Fama and French (1992)).
Quantiles
Low
High
Panel A: Relative Option Volume

OneDayRV (bps)
dIV/IV (bps)
635
7187
4.21
36.36
19.72
31.38
51.10
36.96
4.30
603
6567
4.01
37.05
14.30
44.10
58.40
51.63
70.51
602
7070
3.97
27.02
2.33
40.57
38.24
39.72
16.61
544
6519
2.15
24.85
35.47
2.26
37.73
61.98
28.10
500
6129
0.45
23.80
20.60
161.36
181.96
51.87
39.49
608
7087
2.28
23.25
35.22
0.61
35.83
57.61
0.62
743
8446
5.39
10.81
14.31
64.73
50.42
31.04
7.39
Panel B: Historical Stock Volatility

OneDayRV (bps)
dIV/IV (bps)
356
4438
4.72
40.20
33.00
115.42
148.42
37.14
113.24
386
4791
8.05
42.71
19.42
181.60
201.03
84.72
21.69
519
6343
4.41
28.17
32.99
70.89
103.88
59.86
9.41
(continued )
1075
Table IIContinued
Quantiles
Low
High
Panel C: ln(size)
OneDayRV (bps)
dIV/IV (bps)
746
8276
10.52
13.86
17.41
4.67
22.08
53.97
17.52
697
8124
3.49
26.25
8.51
60.87
52.37
32.60
13.40
548
6691
4.24
25.81
11.58
68.21
79.79
64.67
24.86
479
5964
6.07
29.07
39.98
11.85
28.13
41.21
7.43
391
4899
2.93
32.34
31.16
195.30
226.46
59.84
61.90
477
5748
0.87
26.26
20.31
157.37
137.06
44.62
42.04
591
6972
1.50
26.68
16.43
30.07
46.50
57.39
6.81
Panel D: Book-to-Market Ratio

OneDayRV (bps)
dIV/IV (bps)
585
6737
0.10
21.09
41.16
61.08
102.24
54.27
30.41
556
6531
6.42
31.77
20.50
79.98
100.48
51.02
12.95
523
6308
4.34
30.94
19.75
70.06
89.81
56.78
5.22
in the option market than the stock market. For example, there is a sizeable
increase in option volume but not in stock volume before the EAD. The increased
option volume, however, lasts only until day +1 while the stock volume is still
elevated on day +4. The unique time pattern in the demand for volatility is also
noteworthy. Panel 3 shows that it increases substantially in the days before
the EAD and declines by the EAD. These patterns suggest that volatility and
directional trading behave differently around EADs.
III. Results
A. Information in Option Volume for Future Stock Price Volatility
As detailed in Section I, our first empirical specification is designed to test
whether non-market maker net demand for volatility in the option market
predicts the future volatility of underlying stocks. Specifically, equation (2) is
estimated after filtering observations and constructing variables as described
in Section II. Following Alizadeh, Brandt, and Diebold (2002), the OneDayRV i,t
proxy for the realized volatility of stock i on date t is 10,000 times stock is high
minus low price during day t divided by stock is closing price.16 We perform
16
The results are robust to defining the realized volatility proxy in a number of ways including
simple absolute daily return, the log of the intraday high price minus the log of the intraday low
price, and the log of the difference between the intraday high and low prices.
Table III
Implied volatility (IV)

IV change (dIV)
dIV/IV
Nonstraddle vol. demand
Abs(D )
OptVolume
ln(stkVolume)
0.39
0.14
0.15
0.07
0.09
0.02
0.03
0.00
0.17
0.27
0.47
One
DayRV
0.17
0.17
0.01
0.02
0.02
0.02
0.03
0.08
0.15
0.29
IV
0.97
0.22
0.19
0.07
0.12
0.07
0.03
0.07
0.04
dIV
0.22
0.20
0.07
0.13
0.08
0.04
0.08
0.05
dIV/IV
0.76
0.42
0.19
0.32
0.03
0.01
0.03
Vol. Dem.
0.21
0.17
0.23
0.00
0.06
0.06
Open
Vol. Dem.
0.05
0.14
0.05
0.08
0.04
Close
Vol. Dem.
0.08
0.02
0.01
0.02
Straddle
Vol. Dem.
0.05
0.04
0.00
Nonstraddle
Vol. Dem.
0.55
0.27
Abs(D )
0.44
optVolume
This table reports the correlation coefficients for a number of variables over the January 1990 through December 2001 time period. OneDayRV is 10,000
times the difference of an underlying stocks intraday high and low price divided by the closing stock price. IV is the average implied volatilities of the
selected straddles call and put. The dIV is daily change in IV. Volatility Demand (D ) is the vega-weighted volatility demand defined in equation (1).
Open Volatility Demand, Close Volatility Demand, Straddle Volatility Demand, and Nonstraddle Volatility Demand are volatility demand constructed
only from, respectively, open or close option volume, or option volume that may have been or could not have been part of straddle trades. Abs(D ) is
absolute net delta-weighted option volume, optVolume is total number of option contracts traded, and ln(stkVolume) is log of daily stock volume.
Correlation Coefficients
1076

1. oneDayRV(bps)
800
1077
2. dIV/IV(bps)
200
600
400
200
0
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10
3. D (normalized)
4. abs(D) (normalized)
0.1
0.2
0.05
0.15
0.1
0.05
0
0 1 2 3 4 5 6 7 8 9 10
0.05
5. optVolume (normalized)
6. log(stkVolume) (normalized)
0.15
0.3
0.1
0.2
0.05
0.1
0.05
0.1
10 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10
10 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10
10 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 10
Figure 1. Average value of variables before and after earnings announcement dates. This
figure reports the average value of variables at various numbers of trading days relative to earnings
announcement dates. The horizontal lines correspond to the average value of the variable across the
entire sample. OneDayRV is 10,000 times the difference of an underlying stocks intraday high and
low prices divided by the closing stock price. The dIV/IV is this daily change in the implied volatility
of a short-maturity, close-to-the-money straddle divided by the level of that straddles implied
volatility. This quantity is reported in basis points, consistent with how it is used in the regression
analysis. Volatility demand (D ) is the vega-weighted volatility demand defined in equation (1),
abs(D ) is absolute net delta-weighted option volume, optVolume is total number of option contracts
traded, and ln(stkVolume) is the logarithm of daily stock volume.
pooled regressions and compute t-statistics from robust standard errors that
correct for cross-sectional correlation in the data.
Table IV summarizes the main results. For all values of j = 1, 2, 3, 4, 5 the coefficient on the non-market maker net demand for volatility variable, Di,tj , is
positive and significant, indicating that, indeed, option volume contains information about the future volatility of the underlying stock for at least 5
trading days into the future. The reported coefficient on the interaction term
with the EAD indicator also shows that on the day before the EAD, there is
an additional, significant increase in the informational content of the option
volume.
Table IV also reports the coefficients on the control variables, including the
proxy for the realized volatilities on days t 5 through t 1. Consistent with the
well-known volatility clustering effect from the ARCH/GARCH literature, the
Table IV
Const.
tj
t3
0.03
(4.09)
0.03
(4.70)
0.03
(5.00)
0.03
(4.40)
0.03
(4.90)
t2
t2
Ind
1 2.45 10.42
8.40
0.21 0.04
0.06 0.01 0.04
0.03
(0.89) (11.64) (2.35) (7.96) (1.35) (8.81) (0.43) (5.28) (1.10)
2 2.30
5.95
2.68
0.23 0.06
0.07 0.01 0.04
0.00
(0.88) (8.69) (0.81) (8.97) (1.93) (10.52) (0.71) (5.13) (0.11)
3 2.32
5.07 2.87
0.23 0.05
0.07 0.02 0.04
0.00
(0.88) (7.34) (0.75) (8.81) (1.54) (8.20) (0.75) (5.84) (0.07)
4 2.28
3.40 0.08
0.25 0.05
0.06 0.02 0.04
0.01
(0.90) (5.16) (0.01) (29.29) (2.26) (8.25) (0.96) (5.10) (0.43)
5 2.18
2.13
11.09
0.24 0.05
0.07 0.01 0.04
0.00
(0.84) (3.99) (2.62) (15.22) (1.83) (9.70) (0.50) (4.63) (0.12)
t1
t1
Ind
t4
tj
Ind
OneDayRV
t3
Ind
0.01
(0.33)
0.02
(1.11)
0.01
(0.39)
0.01
(0.23)
0.02
(0.94)
t4
Ind
Ind
t
t1
t
Ind t j
tj
Ind
abs(D )
tj
tj
Ind
OptVolume.
tj
tj
Ind
ln(stkVolume)
Adj.
R2
0.02
0.02 139.52
0.04 0.01 1.24 3.87
0.43
1.43
6.54 7.56 0.15
(4.00) (0.84) (24.12) (19.85) (1.41) (2.16) (1.08) (0.32) (0.26) (2.04) (1.37)
0.02
0.00 136.69
0.04 0.01 0.28 1.07 3.90 8.19
0.18
6.65 0.15
(3.90) (0.14) (23.89) (18.08) (2.65) (0.47) (0.27) (3.07) (1.43) (0.14) (1.18)
0.03
0.03 141.16
0.04 0.01 0.12 2.08 3.26
2.50 3.16 2.89 0.16
(4.88) (1.43) (22.24) (19.57) (2.17) (0.22) (0.50) (2.56) (0.49) (2.49) (0.55)
0.03
0.02 139.26
0.04 0.01 0.80
0.54 4.70
9.12 0.68 8.28 0.16
(4.47) (0.79) (24.39) (26.13) (3.09) (1.56) (0.11) (3.92) (1.60) (0.58) (1.67)
0.03 0.01 134.98
0.04 0.02 1.36 3.69 5.47
4.75 0.32
1.03 0.17
(5.15) (0.52) (23.31) (23.27) (3.13) (2.46) (0.85) (4.96) (0.99) (0.30) (0.22)
t5
t5
Ind
IV
673,615
675,552
678,028
683,212
703,229
Obs
This table reports estimates from pooled regressions over 1990 through 2001. The dependent variable, OneDayRVi,t , is a proxy for the realized volatility of stock i on trade
day t. The proxy is 10,000 times the difference between the stocks intraday high and low price divided by the closing stock price. The net demand variable, D , is the
demand for volatility in the option market for underlying stock i on trade day t j. The IV variable is the implied volatility for the underlying stock on trade day t 1
computed from a short-maturity, close-to-the-money put-call pair. The D variable is the net equivalent number of shares of underlying stock i bought in the option market
on day t j. The optVolume variable is the total number of contracts of options on underlying stock i traded on day t j. The stkVolume variable is the number of shares
of stock i that trade on day t j. The results are presented for j = 1, . . . , 5. The Ind variable is one if trade day t is an earnings announcement day for stock i and zero
otherwise. The second row of the table indicates the time at which that columns variable is measured. The third row of the table indicates whether that columns variable
is interacted with this indicator variable. The parentheses contain t-statistics computed from robust standard errors that correct for cross-sectional correlation in the data.
The Information in Volatility Demand for Future Realized Volatility of the Underlying Stock
1078
1079
coefficients on the lagged realized volatilities are positive and significant. We

also include the lagged implied volatility to control for the information already
contained in the option price. Consistent with the findings in the literature on
the informational content of implied volatility, the coefficients on the IV control
variable are positive and significant. This fact indicates that, not surprisingly,
the implied volatility at the end of day t 1 contains information about the
realized volatility on day t. The inclusion of this control variable also allows us to
conclude that the volatility demand contains information about future volatility
beyond that present in implied volatility. It is commonly believed that implied
volatility impounds all publicly available information about the future volatility
of the underlying stock. Since the market makers can observe volatility demand,
our results suggest that at least some information about future volatility that is
observable to an important set of option market participants is not immediately
ref lected in option prices.17 Finally, the coefficient for abs(D ), which controls
for option market trading on private directional information, is positive but
more often than not insignificant. The comparison between the predictability
of D and abs(D ) is interesting since both demand variables are constructed
from the same set of option trading volumes and are different only in that
D is vega weighted while D is delta weighted. The lack of predictability of
abs(D ) indicates that by weighting the option volume by how sensitive each
option contract is to volatility information, we pick up the volatility-related
information from the option trading volume.
In Table V, we further separate the total non-market maker net demand
for volatility into two groups by taking advantage of the fact that our data
set distinguishes non-market maker volume that opens new option positions
from non-market maker volume that closes existing positions. Panels A and
B of Table V report estimation results when the net demand for volatility is
computed from only open or close option volume, respectively. For the sake of
brevity, Table V does not report the coefficient estimates and t-statistics for the
Ind, lagged OneDayRV, IV, D , optVolume, and ln(stkVolume) control variables.
The omitted coefficient estimates and t-statistics are similar to those reported
in Table IV. The main effect is considerably stronger for open than for close
volume. For example, when j = 1 the coefficient on open volume net volatility
demand is 8.51 (t-statistic = 11.43). On the other hand, when close volume is
used, the corresponding coefficient estimate is just 4.78 (t-statistic = 7.09).
These results are interesting for two reasons. First, investors who trade on
volatility information in the option market are more likely to do so by opening
new option positions; to do so by closing existing positions requires that a trader
who obtains private volatility information has existing option positions on the
underlying stock that the information indicates should be closed. Second, the
fact that the main result is considerably stronger for open than close volume
suggests that microstructure or other mechanical effects such as market maker
17
A caveat here is that implied volatility should subsume all other publicly observable variables
as a forecaster of the realized volatility over the life of the options from which the volatility is
implied. Our tests, however, predict realized volatility only over the next trading day.
1080

Table V
The Information in Open and Close Volatility Demand for Future

Realized Volatility of the Underlying Stock
This table reports estimates from pooled regressions over 1990 through 2001. The dependent variable, OneDayRVi,t , is a proxy for the realized volatility of stock i on trade day t. The proxy is 10,000
times the difference between the stocks intraday high and low price divided by the closing stock
price. The net demand variable, D , is the demand for volatility in the option market for underlying
stock i on trade day t j. In the two panels, it is computed either from open or close option volume.
The IV variable is the implied volatility for the underlying stock on trade day t 1 computed from
a short-maturity, close-to-the-money put-call pair. The D variable is the net equivalent number
of shares of underlying stock i bought in the option market on day t j. The optVolume variable is
the total number of contracts of options on underlying stock i traded on day t j. The stkVolume
variable is the number of shares of stock i that trade on day t j. Each panel presents results for
j = 1, . . . , 5. The Ind variable is one if trade day t is an earnings announcement day for stock i and
zero otherwise. The second row of the table indicates the time at which that columns variable is
measured. The third row of the table indicates whether that columns variable is interacted with
this indicator variable. The parentheses contain t-statistics computed from robust standard errors
that correct for cross-sectional correlation in the data.
D
j
tj
t j Ind
Adj. R2
Obs.
Panel A: Net Volatility Demand from Open Option Volume

1
2
3
4
5
8.51
(11.43)
4.55
(7.19)
4.81
(7.21)
3.13
(5.34)
1.09
(2.03)
6.27
(1.85)
1.01
(0.30)
2.05
(0.49)
0.71
(0.18)
12.05
(2.79)
0.15
703,229
0.14
683,212
0.16
678,028
0.16
675,552
0.16
673,615
Panel B: Net Volatility Demand from Close Option Volume

1
2
3
4
5
4.78
(7.09)
3.10
(5.05)
1.23
(2.46)
1.18
(2.28)
1.80
(3.31)
5.59
(1.58)
7.83
(2.17)
8.11
(2.14)
3.32
(0.40)
3.16
(0.74)
0.15
703,180
0.14
683,164
0.16
677,969
0.16
675,501
0.16
673,570
hedging are unlikely to account for the observed future changes in the volatility
of the underlying stocks.18 For example, market makers will hedge their new
18
Since market makers nearly always hedge their option positions on the same day that they
are entered, any volatility in the underlying stock caused by hedging should be concentrated on
day j = 0 anyway.
1081
option positions in the same way regardless of whether they result from nonmarket makers opening or closing option positions. For both of these reasons,
the stronger findings for open than close volume provide evidence that our
main result is driven by investors bringing private volatility information to the
market.
We generate further evidence on whether volatility information trading lies
behind our results by identifying option volume that is likely to have higher and
lower concentrations of informed volatility trading. Some option market trades
represent bets on both the level and the volatility of the underlying stock. For
example, an investor who buys a naked call benefits from an increase in the
level of the underlying stock price and also from an increase in its volatility.
Other trades, however, bet primarily on increases or decreases in volatility,
and, presumably, informed volatility traders would tend to make such trades.
The leading example of these trades is a straddle where an investor buys or
sells a call and a put with the same strike price and time to expiration. If
the predictability that we identify in the option volume indeed has its source
in informed volatility trading, then the predictability should be stronger from
volume that has a higher concentration of straddle trading.
We test this hypothesis by separating the option volume into that that could
have been part of straddle trades and that that could not have been part of
straddles. We define volume that could have been part of a straddle trade to
be matching call and put volume on a given trading day. In order for call and
put volume to be matching, it must be on the same underlying stock with the
same strike price and time to expiration. In addition, it must be from the same
class of investor (i.e., firm proprietary traders, discount customers, full service
customers, or other public customers) and match on whether it is buy or sell
volume and also on whether it is open or close volume. Finally, when the number
of contracts of call and put volume that meet these criteria is not the same, the
smaller number of contracts of both the call and put volume are counted as
potentially part of straddle trades. All volume that is not potentially part of
straddle trades is defined as volume that could not have been part of straddle
trades.19
Table VI reports the respective predictability when the net volatility demand
is computed from volume that could have been part of straddles and volume
that could not have been part of straddles. Again, all controls specified in equation (2) are included in the estimation but not reported in the table. For all
five values of j, the predictability is stronger when demand is computed from
19
In order to illustrate our classification, suppose that on a given trade day there are five
contracts of firm proprietary trader open buy IBM call volume with strike price 20 and expiration
next month, and eight contracts of firm proprietary open buy IBM put volume with strike price 20
and expiration next month. In this case, all five contracts of the call volume and five of the eight
contracts of the put volume will be classified as potentially part of straddle trades while the other
three contracts of put volume will be classified as not part of straddle trades. Although not all
volume classified as potentially part of straddle trades is in fact part of straddle trades (e.g., if the
call and put volume comes from different investors within the same investor class), this volume
certainly has a higher concentration of straddles than the volume classified as not part of straddles
trades because the latter contains no straddle trading at all.
1082

Table VI
The Information in Potential Straddle and Nonstraddle Volatility

Demand for Future Realized Volatility of the Underlying
Stock Volume
times the difference between the stocks intraday high and low price divided by the stocks closing
stock i on trade day t j. In addition to the variables reported in the table, OneDayRV for t 1
through t 5, IV, D , optVolume, and ln(stkVolume) both by themselves and interacted with the
Ind indicator variable that is one if day t is an earnings announcement day for stock i are included
as controls. The Ind indicator variable by itself is also included as a control. The second row of
the table indicates the time at which that columns variable is measured. The third row of the
table indicates whether that columns variable is interacted with this indicator variable. Panels A
and B contain results from running the regressions where the net demand variable, D , is constructed from, respectively, (1) only option volume that could have been part of straddles, and (2)
only option volume that could not have been part of straddles. Each panel presents results for j =
1, . . . , 5. The parentheses contain t-statistics computed from robust standard errors that correct for
cross-sectional correlation in the data.
D
j
tj
t j Ind
Adj. R2
Obs.
Panel A: Volatility Demand from Volume That Could Have Been Part of Straddles
1
2
3
4
5
9.18
(8.94)
3.77
(5.15)
3.19
(4.79)
1.88
(2.69)
1.55
(2.12)
10.09
(1.99)
4.63
(0.80)
0.13
(0.02)
9.06
(0.65)
2.01
(0.32)
0.19
268,385
0.19
268,422
0.19
268,436
0.19
268,440
0.19
268,424
Panel B: Volatility Demand from Volume That Could Not Have Been Part of Straddles
1
2
3
4
5
3.63
(4.98)
1.73
(2.63)
0.87
(1.45)
0.67
(0.95)
0.41
(0.62)
14.13
(2.69)
5.35
(1.08)
0.91
(0.14)
0.39
(0.07)
10.48
(1.72)
0.19
268,385
0.19
268,422
0.19
268,436
0.19
268,440
0.19
268,424
volume that could have been part of straddles than when demand is computed from volume that could not have been part of straddles. For example,
when j = 1, the coefficient estimate on D is 9.18 (t-statistic = 8.94) when demand is computed from volume that could have been part of straddles and 3.63
1083
(t-statistic = 4.98) when computed from volume that could not have been part
of straddles. The stronger results from volume that is potentially part of straddles support the proposition that the predictability has its source in volatility
information trading.
As discussed in Section I, it is important to examine whether the positive relation between non-market maker net demand for volatility and future realized
volatility is driven by option market trading on private directional information.
For this reason, the specification in equation (2) includes the D variables to
control for directional trading. To further investigate whether option market
trading on directional information is likely to be an important factor in the
previous results, we rerun the tests from Table IV separately on underlying
stocks that have open buy volume put-call ratios in the bottom, middle, and top
terciles. That is, on each trading day we assign observations to terciles based
on the open buy volume put-call ratio of the underlying stocks and then run
the pooled regression separately on the observations that fall into each of the
terciles. We run these tests because Pan and Poteshman (2006) show that the
open buy volume put-call ratio is a strong predictor of the direction of future
movements of underlying stock prices. If investors bringing directional information to the option market play an important role in generating the results
reported in Table IV, then we would expect the findings to be weaker for the
middle tercile observations (for which there is relatively little directional information in the option volume) and stronger in the low (high) tercile observations,
which contain a good deal of information on whether stock prices are going to
subsequently increase (decrease).
The results for the three open buy volume put-call ratio terciles are reported
in Table VII. As in Pan and Poteshman (2006), stock trading dates are included
in these regressions when there are at least 50 contracts of open buy option
volume. For all five values of j the middle put-call ratio tercile regression has
the largest coefficient on the net demand variable. Hence, there is no evidence
that the relation between non-market maker net demand for volatility and
future realized volatility is weaker when there is less directional information
in the option volume. Indeed, it appears that the relation may be stronger
when there is less directional information. We conclude that it is unlikely that
our evidence that option volume is predictive for future volatility is driven by
directional predictability in the option volume.
Finally, we examine the cross-sectional variation in the volatility demand
predictability using observable characteristics of the underlying stock. Specifically, we add to the specification in equation (2) the following cross-sectional
variables, which were defined in Subsection II.C: relOptVolume (the ratio of
option volume to stock volume), histStockVol (the annual realized stock volatility), ln(size) (the logarithm of firm size), and BM (the book-to-market ratio). We
also add their interactions with volatility demand to the predictive analysis so
that we can examine whether the predictability that we document varies across
different firm characteristics.
The results are presented in Table VIII, which again for the sake of brevity
omits the coefficient estimates and t-statistics for the control variables. The
1084

Table VII
The Information in Volatility Demand for Future Realized Volatility

of the Underlying Stock: Underlying Stocks with Different Open Buy
Put-Call Ratios
This table reports estimates from pooled regressions over 1990 through 2001. The dependent variable,
OneDayRVi,t , is a proxy for the realized volatility of stock i on trade day t. The proxy is 10,000 times the
difference between the stocks intraday high and low price divided by the stocks closing price. The net demand
variable, D , is the demand for volatility in the option market for underlying stock i on trade day t j. In
addition to the variables reported in the table, OneDayRV for t 1 through t 5, IV, D , optVolume, and
ln(stkVolume) are included as controls both by themselves and interacted with the Ind indicator variable
that is one if day t is an earnings announcement day for stock i. The Ind indicator variable is also included
as a control. The second row of the table indicates the time at which that columns variable is measured. The
third row of the table indicates whether that columns variable is interacted with this indicator variable.
Panels A through C contain the results from running the regressions on observations with underlying stocks
in the lowest (1), middle (2), and highest (3) terciles of daily open buy volume put-call ratio for underlying
stocks, and each panel presents results for j = 1, . . . , 5. The parentheses contain t-statistics computed from
robust standard errors that correct for cross-sectional correlation in the data.
D
j
tj
t j Ind
Adj. R2
Obs.
0.17
116,576
0.17
113,946
0.17
113,041
0.17
112,517
0.17
112,292
0.16
188,017
0.16
185,857
0.16
185,029
0.16
184,541
0.18
184,190
0.16
184,800
0.17
180,865
0.18
179,709
0.18
179,326
0.16
178,748
Panel A: Low PC Tercile

1
2
3
4
5
8.96
(8.92)
4.57
(5.42)
4.50
(5.25)
3.52
(4.03)
2.12
(2.92)
3.31
(0.51)
4.97
(0.80)
8.58
(1.38)
10.86
(1.39)
8.65
(1.27)
Panel B: Middle PC Tercile
1
2
3
4
5
12.46
(7.96)
6.64
(6.20)
4.94
(4.31)
3.84
(3.76)
2.56
(2.98)
9.08
(1.27)
4.94
(0.79)
6.81
(0.89)
2.83
(0.36)
18.36
(2.17)
Panel C: High PC Tercile
1
2
3
4
5
8.17
(9.12)
3.89
(5.50)
4.45
(5.30)
2.00
(3.06)
0.78
(1.07)
11.75
(1.82)
0.87
(0.16)
0.54
(0.07)
3.77
(0.28)
15.78
(1.90)
1085
Table VIII
Cross-sectional Heterogeneity in the Information in Volatility

Demand for Future Realized Volatility of the Underlying Stock
times the difference between the stocks intraday high and low price divided by the closing stock
stock i on trade day t j. The D measure appears alone and is also interacted with an indicator
variable and four stock characteristics: relOptVolume (100 times the ratio of option volume to stock
volume over the previous calendar year), histStockVol (100 times the sample standard deviation of a
stock over the past calendar year), ln(size) (the natural logarithm of the market capitalization of the
stock at the end of the previous calendar year), and BM (the stocks book-to-market ratio as in Fama
and French (1992)). In addition to the variables reported in the table, OneDayRV for t 1 through
t 5, IV, D , optVolume, and ln(stkVolume) are included as controls both by themselves and interacted with the Ind indicator variable. The uninteracted Ind indicator, relOptVolume, histStockVol,
ln(size), and BM variables are also included as controls. The Ind variable is one if trade day t is
an earnings announcement day for stock i and zero otherwise. The second row of the table indicates the time at which that columns variable is measured. The third row of the table indicates
whether that columns variable is interacted with the indicator variable or a stock characteristic
variable. The parentheses contain t-statistics computed from robust standard errors that correct
for cross-sectional correlation in the data.
D
j
tj
tj
Ind
tj
relOptVolume
tj
histStockVol
tj
ln(size)
tj
BM
Adj.
R2
Obs.
10.87
(3.62)
8.40
(4.84)
5.18
(3.60)
3.07
(1.60)
3.40
(2.61)
8.71
(2.45)
2.71
(0.82)
3.00
(0.78)
0.00
(0.00)
11.17
(2.64)
6.00
(1.85)
0.60
(0.18)
4.51
(1.39)
4.69
(1.49)
2.38
(0.86)
1.28
(1.82)
0.27
(0.73)
0.77
(2.01)
0.66
(1.59)
0.07
(0.23)
1.69
(5.14)
1.40
(5.00)
1.10
(4.01)
0.85
(2.80)
0.27
(1.05)
7.67
(4.98)
5.96
(4.51)
3.12
(2.47)
1.19
(1.00)
1.76
(1.52)
0.15
703,229
0.15
683,212
0.16
678,028
0.16
675,552
0.16
673,615
2
3
4
5
coefficient on the term that interacts D with ln(size) is negative for all values
of j and statistically significant for j = 1, 2, 3, and 4. The negative coefficient
estimate indicates that there is less private volatility information per unit of
volatility demand for larger stocks. This result is consistent with the idea that
information f low is more efficient for larger stocks. The BM interaction term
also has all negative coefficient estimates that are significant for j = 1, 2, and
3. This finding suggests that investors bring more volatility information to the
option market for glamour firms. The estimates for the relOptVolume and histStockVol terms are mostly insignificant. The relOptVolume estimates are all
negative. To the extent that these insignificant negative signs have any meaning, they are consistent with there being greater volatility information per unit
of volatility demand for underlying stocks that have less option market relative to stock market trading. This may be because stocks with relatively more
1086
option market trading have a higher baseline of option market activity unrelated to private volatility information. The point estimates for the histStockVol
interaction term are mostly positive. If the sign of these estimates actually is
meaningful despite their lack of statistical significance, it would indicate that
there is a higher concentration of option trading based on private volatility
information on higher volatility stocks.
B. Price Impact and Informational Asymmetry
The previous subsection demonstrates that non-market maker net demand
for volatility predicts the future volatility of underlying stocks. This fact was
interpreted as evidence that investors use the option market to trade on volatility information. It is therefore natural to explore the pricing consequences of
investors bringing volatility information to the option market. We next investigate whether option prices are impacted by market makers protecting themselves from investors who possess private information about the future volatility of underlying stocks.
The results of estimating equation (4) are reported in Table IX. Once again,
we perform pooled regressions and compute t-statistics from robust standard
errors that correct for cross-sectional correlation in the data. The dependent
variable is the daily percent change in the implied volatility in units of basis
points. As reported in the first column of Table IX, the coefficient on the nonmarket maker net demand for volatility variable is 154 basis points and is
highly significant, implying that on average a one-standard deviation increase
in volatility demand increases implied volatility by over 1.5%.
In order to investigate cross-sectional heterogeneity in the price impact of
volatility demand, specification (4) includes the four underlying stock characteristic variables along with their interactions with the demand variable. The
only one of these terms that has a significant coefficient estimate is the one
that interacts volatility demand with firm size. The negative coefficient on this
interaction term indicates that the price impact per unit of volatility demand
is greater for smaller stocks. Once again, this result is consistent with smaller
firms having less efficient information f low.
As discussed in Subsection I.B, the positive and significant price impact coefficient on the volatility demand variable could be driven by two components:
demand pressure and informational asymmetry. We investigate in two ways
whether informational asymmetry plays an important role in producing the
price impact. Both approaches take advantage of the variation in informational
asymmetry across time and across different kinds of option trading volume.
First, we examine the price impact of a unit of volatility demand as information asymmetry increases leading up to EADs. The coefficients on the interaction terms D Ind(EAD + n) give the incremental price impact of a unit of
volatility demand associated with it being day n relative to an EAD. As shown
in the first column of Table IX, in the 5 days leading up to the EAD (i.e., when
n = 5, 4, 3, 2, and 1), the marginal impact of a unit of volatility demand
is positive with a coefficient ranging between 23 and 63. Furthermore, the
1087
Table IX
Price Impact per Unit of Volatility Demand

This table reports estimates from pooled regressions over 1990 through 2001. The dependent variable
is 10,000 times the daily change in implied volatility divided by the level of the implied volatility
for underlying stock i on trade date t 1. The D variable is the net non-market maker demand for
volatility in the option market for underlying stock i on trade day t. D,Open is the same computed
only from option volume that opens new option positions. relOptVolume is 100 times the option volume
divided by the stock volume, histStockVol is 100 times the standard deviation of daily returns from the
past year, size is market capitalization, and BM is the book-to-market ratio. The Ind(EAD n) variable
is one if trade day t + n is an earnings announcement for stock i and zero otherwise. The parentheses
contain t-statistics computed from robust standard errors that correct for cross-sectional correlation
in the data.
Intercept
D
(D )2
D,Open
D relOptVolume
D histStockVol
D ln (size)
D BM
relOptVolume
histStockVol
ln(size)
BM
D Ind(EAD 5)
D Ind(EAD 4)
D Ind(EAD 3)
D Ind(EAD 2)
D Ind(EAD 1)
D Ind(EAD 0)
D Ind(EAD + 1)
D Ind(EAD + 2)
D Ind(EAD + 3)
D Ind(EAD + 4)
D Ind(EAD + 5)
Ind(EAD 5)
Ind(EAD 4)
Ind(EAD 3)
Ind(EAD 2)
Ind(EAD 1)
Ind(EAD 0)
Ind(EAD + 1)
Ind(EAD + 2)
Ind(EAD + 3)
Ind(EAD + 4)
Ind(EAD + 5)
Adj. R2
Obs.
8.31
153.82
(1.81)
(31.55)
8.36
151.29
11.85
(1.75)
(27.52)
(5.87)
2.58
0.48
15.96
3.85
0.82
0.14
0.47
1.03
23.02
33.34
22.56
42.32
62.89
52.5
26.94
11.37
3.31
13.3
6.68
47.4
32.8
64.74
85.31
109
80.54
338.91
78.84
63.84
59.68
48.86
(1.38)
(0.55)
(13.84)
(0.72)
(0.46)
(0.14)
(0.38)
(0.21)
(1.42)
(2.25)
(1.47)
(2.77)
(4.16)
(3.90)
(1.87)
(1.04)
(0.21)
(0.60)
(0.46)
(4.24)
(2.55)
(5.32)
(6.50)
(9.01)
(5.62)
(23.03)
(7.05)
(5.72)
(4.40)
(4.52)
2.00
0.56
15.93
0.00
0.91
0.08
0.37
0.00
23.85
31.48
28.66
42.22
63.19
53.77
29.75
9.72
1.14
12.95
5.25
48.44
33.09
63.06
85.40
108.87
81.06
342.80
80.44
64.02
59.69
49.17
(1.06)
(0.61)
(11.73)
(5.20)
(0.51)
(0.08)
(0.28)
(0.08)
(1.49)
(2.11)
(1.76)
(2.76)
(4.23)
(4.00)
(2.02)
(0.87)
(0.07)
(0.58)
(0.37)
(4.33)
(2.58)
(5.12)
(6.53)
(9.04)
(5.66)
(23.17)
(7.18)
(5.76)
(4.38)
(4.54)
0.04
238,509
0.04
238,509
8.40
125.10
(1.76)
(21.22)
30.57
1.46
0.80
15.02
0.00
0.86
0.07
0.37
0.00
22.27
29.81
25.92
41.12
61.35
51.66
29.33
9.07
2.12
10.88
7.46
46.92
31.53
61.84
84.63
107.23
82.95
343.42
79.50
63.72
58.77
49.58
(10.74)
(0.77)
(0.89)
(11.14)
(5.01)
(0.48)
(0.07)
(0.28)
(0.08)
(1.39)
(2.00)
(1.57)
(2.70)
(4.09)
(3.85)
(2.04)
(0.83)
(0.13)
(0.49)
(0.52)
(4.21)
(2.46)
(5.02)
(6.47)
(8.91)
(5.79)
(23.35)
(7.13)
(5.77)
(4.31)
(4.60)
0.04
238,509
coefficient estimates are close to monotonically increasing in the week leading up to the EAD, and the estimates for n = 2 and 1 are significant at
conventional levels. These coefficient estimates indicate that as informational
asymmetry increases in the days leading up to EADs, the price impact of a fixed
1088
quantity of volatility demand also increases. These results are consistent with
market makers protecting themselves from greater concentrations of volatility
information being brought to the option market as EADs approach. Indeed, the
point estimate of 63 basis points for day 1 indicates that the price impact per
unit of volatility demand increases by over 40% (= 62.89/153.82) from its baseline level just before EADs. This pattern of time variation in the price impact is
a strong indication that informational asymmetry is an important component
of the price impact captured in our paper. If the demand pressure were the main
driving force, then we would not expect to observe this large variation in price
impact.
As can be seen in Panel 3 of Figure 1, volatility demand pressure tends to
be higher leading up to EADs. One might be concerned that the price impact
of demand pressure is nonlinear with greater than linear increases in impact
when volatility demand is high. To investigate this possibility, we add a squared
volatility demand variable to the regression. The third and fourth columns of
Table IX indicate that adding this variable to the regression results in hardly
any change to the magnitude or significance of the coefficient estimates on
the terms that interact demand for volatility with indicators of the number of
trading days relative to an EAD.
The second approach to examining whether informational asymmetry plays
a role in the price impact takes advantage of the fact that our data set distinguishes non-market maker volume that opens new option positions from
non-market maker volume that closes existing positions. Indeed, as reported
above in the predictive analysis, the volatility demand constructed from the
open volume has a higher predictability than that from the close volume. For
this reason, we included a second demand variable, D,Open , in our price impact
analysis. With the total demand variable in the same regression, the slope coefficient on D,Open effectively tests the differential price impact of the volatility
demand constructed from open and close volume. Columns 5 and 6 of Table IX
report that the coefficient estimate on this variable is positive and significant,
indicating a greater price impact from the open volume demand. Clearly, pure
demand pressure cannot explain this result. On the other hand, this empirical
fact is consistent with the informational asymmetry explanation provided that
market markers have some ability to distinguish open transactions from close
transactions. Market makers are likely to develop some ability to distinguish
between open and close volume in real time, because at the end of each trade
day they can infer the net open and close volume on each contract from changes
in open interest. Insofar as it is more difficult for them to develop this ability, it will bias against the positive and significant coefficient estimate that is
observed.
Motivated by the same intuition, we also separate option volume by the
moneyness of the option contract and construct the corresponding volatility
demand variables. We find that the price impact per unit of net volatility
demand is strongest for options that are very near the money, less strong
but still appreciable for out-of-the-money options, and weaker but still significant for in-the-money options. These findings are not surprising, because
1089
near-the-money options have the greatest price sensitivity to volatility. Consequently, investors with volatility information are more likely to trade near-themoney options, and market makers adjust option prices most aggressively in
response to near-the-money net demand for volatility.20
Finally, the control variables in Table IX include the indicator variables
Ind(EAD + n) to control for any overall changes in implied volatility around
EADs that are not related to volatility demand. The large positive coefficient
estimates on these indicator variables on days 2 and 1 and the negative
coefficient estimates on days + 1 and + 2 relative to the EAD indicate that, not
surprisingly, implied volatility rises before EADs and falls following them.
IV. Conclusion
Options play several important economic roles including that of providing
a mechanism for investors to bring information to the financial markets. Investors can trade on directional information in either the stock or the option
market, and a number of papers have investigated whether option volume is
informative about the future direction of underlying stock prices. However,
while the option market is uniquely suited for trading on volatility information, there is little research on whether option volume is informative about the
future volatility of underlying stocks.
This paper provides evidence that option volume is informative about future volatility by showing that non-market maker net demand for volatility in
the option market is positively related to the subsequent realized volatility of
underlying stocks, even after controlling for the public information about future volatility impounded in option-implied volatility. A natural interpretation
of this finding is that investors choose to trade on private volatility information in the option market. This interpretation is supported by our findings that
volatility demand from transactions that open new option positions is a stronger
predictor of future volatility than demand that closes existing option positions
and that demand from transactions that could have been part of straddle trades
is a stronger predictor than demand that could not have been part of straddle
trades.
We also investigate the pricing implications of volatility information trading
in the option market. We find that non-market maker net demand for volatility
impacts the prices of options and that the impact increases when informational
asymmetry increases in the days leading up to earnings announcements. This
finding is consistent with option market makers changing prices in order to
protect themselves from investors with volatility information and with their
becoming more concerned about protecting themselves at times when informational asymmetry is especially high.
20
For the sake of brevity, we do not report these results. This finding also provides an interesting
contrast to the findings for option trading on directional information. Black (1975) hypothesizes and
Easley, OHara, and Srinivas (1998) show theoretically that investors with directional information
on underlying stocks will use more highly out-of-the-money contracts to trade in the option market.
Pan and Poteshman (2006) provide empirical support for these predictions.
1090

REFERENCES
Alizadeh, Sassan, Michael W. Brandt, and Francis X. Diebold, 2002, Range-based estimation of
stochastic volatility models, Journal of Finance 57, 10471091.
Amin, Kaushik I., and Charles M. C. Lee, 1997, Option trading, price discovery, and earnings news
dissemination, Contemporary Accounting Research 14, 153192.
Arrow, Kenneth J., 1964, The role of securities in the optimal allocation of risk-bearing, Review of
Economics Studies 31, 9196.
Back, Kerry, 1993, Asymmetric information and options, Review of Financial Studies 6, 435472.
Black, Fisher, 1975, Fact and fantasy in the use of options, Financial Analysts Journal 31, 3641,
6172.
Bollen, Nicholas P., and Robert E. Whaley, 2004, Does net buying pressure affect the shape of
implied volatility functions? Journal of Finance 59, 711753.
Canina, Linda, and Stephen Figlewski, 1993, The informational content of implied volatility, Review of Financial Studies 6, 659681.
Cao, Charles, Zhiwu Chen, and John M. Griffin, 2005, Information content of option volume prior
to takeovers, Journal of Business 78, 10731109.
Chakravarty, Sugato, Huseyin Gulen, and Stewart Mayhew, 2004, Informed trading in stock and
option markets, Journal of Finance 59, 12351257.
Chan, Kalok, Y. Peter Chung, and Wai Ming Fong, 2002, The informational role of stock and option
volume, Review of Financial Studies 15, 10491075.
Chernov, Mikhail, 2002, On the role of volatility risk premia in implied volatilities based forecasting
regressions, Working paper, Columbia University.
Christensen, Bent Jesper, and Nagpurnanand Prabhala, 1998, The relation between implied and
realized volatility, Journal of Financial Economics 50, 125150.
Day, Theodore, and Craig Lewis, 1992, Stock market volatility and the information content of stock
index options, Journal of Econometrics 52, 267287.
Donders, Monique, Roy Kouwenberg, and Ton Vorst, 2000, Options and earnings announcements:
An empirical study of volatility, trading volume, open interest, and liquidity, European Financial Management 6, 149172.
Dubinsky, Andrew, and Michael Johannes, 2005, Earnings announcements and equity options,
Working paper, Columbia University.
Easley, David, Maureen OHara, and P. Srinivas, 1998, Option volume and stock prices: Evidence
on where informed traders trade, Journal of Finance 53, 431465.
Ederington, Louis, and Wei Guan, 2002, Is implied volatility an informationally efficient and effective predictor of future volatility? Journal of Risk 4, 2946.
Ederington, Louis, and Jae Ha Lee, 1996, The creation and resolution of market uncertainty: The
impact of information releases on implied volatility, Journal of Financial and Quantitative
Analysis 31, 513539.
Fama, Gene, and Kenneth French, 1992, The cross-section of expected stock returns, Journal of
Finance 47, 427465.
Garleanu,
Nicolae, Lasse H. Pedersen, and Allen M. Poteshman, 2006, Demand-based option pricing, Working paper, University of Illinois at Urbana-Champaign.
Grossman, Sanford J., 1977, The existence of futures markets, noisy rational expectations and
information externalities, Review of Economic Studies 44, 431449.
Hadi, Mohammed, 2006a, Traders position for action in ford, kinetic concepts, Wall Street Journal
July 19, C6.
Hadi, Mohammed, 2006b, H-Ps puts see heavy trading after Dell warning, Wall Street Journal
July 22, B6.
Hadi, Mohammed, 2006c, Starbucks sets trading pace; Neurocrine volatility heats up, Wall Street
Journal September 1, C4.
Hadi, Mohammed, 2006d, Sepracor, Palm, Fedex calls active, Dow Jones Newswires September 20,
15, 35.
Hadi, Mohammed, 2006e, Traders place bets on Harrahs next move, Dow Jones Newswires October
2, 15, 30.
1091
Hadi, Mohammed, 2006f, McDonalds trades heavily, Wall Street Journal October 5, C5.
Harvey, Campbell R., and Robert E. Whaley, 1992, Market volatility prediction and the efficiency
of the S&P 100 Index option market, Journal of Financial Economics 31, 4373.
Jorion, Philippe, 1995, Predicting volatility in the foreign exchange market, Journal of Finance 50,
507528.
Kyle, Albert S., 1985, Continuous auctions and insider trading, Econometrica 53, 13151336.
Lakonishok, Josef, Inmoo Lee, Neil D. Pearson, and Allen M. Poteshman, 2007, Option market
activity, Review of Financial Studies 20, 813857.
Lamoureux, Christopher, and William Lastrapes, 1993, Forecasting stock-return variance: Toward
an understanding of stochastic implied volatilities, Review of Financial Studies 6, 293326.
Latane, Henry A., and Richard J. Rendleman, 1976, Standard deviations of stock price ratios
implied in option prices, Journal of Finance 31, 369381.
Lee, Charles, and Mark Ready, 1991, Inferring trade direction from intraday data, Journal of
Finance 46, 733746.
Pan, Jun, and Allen M. Poteshman, 2006, The information in option volume for future stock prices,
Review of Financial Studies 19, 871908.
Patell, James M., and Mark A. Wolfson, 1979, Anticipated information releases ref lected in call
option prices, Journal of Accounting and Economics 1, 117140.
Patell, James M., and Mark A. Wolfson, 1981, The ex ante and ex post price effects of quarterly
earnings announcements ref lection in option and stock prices, Journal of Accounting Research
19, 434458.
Pearson, Neil D., Allen M. Poteshman, and Joshua White, 2007, Does option trading have a pervasive impact on underlying stock prices? Working paper, University of Illinois at UrbanaChampaign.
Poteshman, Allen M., 2000, Forecasting future volatility from option prices, Working paper, University of Illinois at Urbana-Champaign.
Ross, Stephen A., 1976, The arbitrage theory of capital asset pricing, Journal of Economic Theory
13, 341360.
Savickas, Robert, and Arthur Wilson, 2003, On inferring the direction of option trades, Journal of
Financial and Quantitative Analysis 38, 881902.
Stephan, Jens A., and Robert Whaley, 1990, Intraday price change and trading volume relations
in the stock and stock option markets, Journal of Finance 45, 191220.
Stulz, Rene M., 2004, Should we fear derivatives? Journal of Economic Perspectives 18, 173192.
Whaley, Robert, and Joseph Cheung, 1982, Anticipation of quarterly earnings announcements: A
test of option market efficiency, Journal of Accounting and Economics 4, 5783.

Volatility Information Trading in The Option Market

Uploaded by

Copyright:

Available Formats

Volatility Information Trading in The Option Market

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Volatility Information Trading in The Option Market

Uploaded by

Copyright:

Available Formats

THE JOURNAL OF FINANCE VOL. LXIII, NO.

Volatility Information Trading

The Journal of Finance

Volatility Information Trading in the Option Market

This interpretation is corroborated by two additional results. First, we find

The Journal of Finance

Volatility Information Trading in the Option Market

are also related to the findings of Garleanu,

The Journal of Finance

Volatility Information Trading in the Option Market

60 trading days of returns leading up to t. It should be noted that our measure

OneDayRV i,t = a + b Di,t

+ d OneDayRV i,t1 + e OneDayRV i,t1 Indi,t

The Journal of Finance

Volatility Information Trading in the Option Market

The Journal of Finance

demand. The economic interpretation of this price impact variable could be

Volatility Information Trading in the Option Market

for = 5, . . . , 1, 0, 1, . . . , 5, therefore captures the incremental price impact

The Journal of Finance

Volatility Information Trading in the Option Market

Volatility Information Trading in the Option Market

straddle trades. Descriptive statistics for the demand variable computed in

The Journal of Finance

Panels 3 through 6 present information on the trading patterns in the stock

Summary Statistics Broken Down by Stock Characteristics

Panel A: Relative Option Volume

Panel B: Historical Stock Volatility

Volatility Information Trading in the Option Market

Panel D: Book-to-Market Ratio

Implied volatility (IV)

Volatility Information Trading in the Option Market

Volatility Information Trading in the Option Market

coefficients on the lagged realized volatilities are positive and significant. We

The Journal of Finance

The Information in Open and Close Volatility Demand for Future

Panel A: Net Volatility Demand from Open Option Volume

Panel B: Net Volatility Demand from Close Option Volume

Volatility Information Trading in the Option Market

The Journal of Finance

The Information in Potential Straddle and Nonstraddle Volatility

Volatility Information Trading in the Option Market

The Journal of Finance

The Information in Volatility Demand for Future Realized Volatility

Panel A: Low PC Tercile

Volatility Information Trading in the Option Market

Cross-sectional Heterogeneity in the Information in Volatility

The Journal of Finance

Volatility Information Trading in the Option Market

Price Impact per Unit of Volatility Demand

The Journal of Finance

Volatility Information Trading in the Option Market

The Journal of Finance

Volatility Information Trading in the Option Market

You might also like