the informational role of option trading volume in the s&p 500 futures options markets
TRANSCRIPT
This article was downloaded by: [Florida International University]On: 20 December 2014, At: 05:45Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Applied Financial EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rafe20
The informational role of option trading volume in theS&P 500 futures options marketsGhulam Sarwara Department of Finance , Insurance and Real Estate, St. Cloud State University , St. Cloud,MN 56301-4498 E-mail:Published online: 02 Feb 2007.
To cite this article: Ghulam Sarwar (2004) The informational role of option trading volume in the S&P 500 futures optionsmarkets, Applied Financial Economics, 14:16, 1197-1210, DOI: 10.1080/0960310042000280483
To link to this article: http://dx.doi.org/10.1080/0960310042000280483
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
The informational role of option trading
volume in the S&P 500 futures options
markets
GHULAM SARWAR
Department of Finance, Insurance and Real Estate, St. Cloud State University,St. Cloud, MN 56301-4498E-mail: [email protected]
This paper analyses the dynamic relations between future price volatility of theS&P 500 index futures and trading volume of S&P 500 futures options toexamine the informational role of the option volume in predicting the futureprice volatility. Using a pooled cross-sectional and time-series data framework,the paper uses the error components and dummy variable models to allow forthe relations between volatility and volume to vary by the option’s time-to-maturityand moneyness. The results suggest that previous call and put volumes have astrong predictive ability with respect to the future price volatility. The results alsoindicate that the future price volatility has a leading positive effect on the optionvolume, but that the rises and falls in volatility exert asymmetric influences on theoption volume. These findings support the hypothesis that both the information-and hedge-related trading explain most of the trading volume of S&P 500 futuresoptions.
I . INTRODUCTION
Options on equity index futures are widely used for
hedging the risk of equity positions and for speculativepurposes. Several recent studies investigate whether the
trading volume of equity options contain information
about the future stock price movements (Easley et al.,
1998; Llorence et al., 2002; Chan et al., 2002). Black(1975) and Mayhew et al. (1995) argue that lower transac-
tion costs and greater financial leverage of the options
markets may induce informed traders to trade in the
option market rather than in the stock market. Further,Back (1993) and Cherian (1993) argue that investors
who possess private information about the future
volatility of the stock price may be more attracted to
the option market instead of the stock market because
they can only make their bet on volatility in the option
market.Easley et al. (1998) examine the lead–lag relations
between the trading volume of options on individual stocks
and the prices of the underlying stocks, and show that theoption trading volume indeed contains information about
the future stock prices. However, Chan et al. (2002) report
that the option net trade volume has no strong predictiveability for stock quote revisions. Similarly, many studies
examine the links between option prices and stock prices,option volume and stock volume, option prices and stock
volume, and stock prices and stock volume (Manaster
and Rendleman, 1982; Bhatacharya, 1987; Anthony,1988; Vijh, 1988, 1990; Stephen and Whaley, 1990; Chan
et al., 1993; Sheikh and Ronn, 1994; Mayhew et al., 1995;
Fleming et al., 1996; Llorence et al., 2002). However, none
Applied Financial Economics ISSN 0960–3107 print/ISSN 1466–4305 online # 2004 Taylor & Francis Ltd 1197
http://www.tandf.co.uk/journalsDOI: 10.1080/0960310042000280483
Applied Financial Economics, 2004, 14, 1197–1210
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
of these studies examine the direct links between the option
trading volume and the stock price volatility expected in
the future. The ability to predict price volatility is impor-
tant for portfolio selection and asset management as well as
for the pricing of primary and derivative securities (Engle
and Ng, 1993). Because informed traders can make direct
bets on their private information of the future price
volatility only in the option market (Back, 1993), an
examination of the interrelations between the future price
volatility and the option trading volume could provide use-
ful insights about the volatility-related information content
of option trades. The option pricing theory is unclear about
the exact nature of such volume–volatility relations.
Because the hedging and speculative uses of options arise
from the asset’s price volatility, some researchers may
argue that the option trading volume should unilaterally
follow the price volatility. For example, a perceived or an
unexpected actual increase in the stock price volatility
could lead to a higher trading volume of stock options as
traders increase their risk management and speculative
activities to cut the potential losses or to enhance the
potential gains on their equity positions. In fact, Sears
(2000a, b) and Tan (2001a) cite a positive relation between
the higher perceived price volatility and the trading volume
of stock options. However, the argument for the unidirec-
tional link from the price volatility to the option trading
volume implicitly assumes perfect markets and symmetric
information across traders, and views the trading volume
as a non-informative outcome of the trading process.
If some traders are better informed than others, and if
traders cannot accurately forecast the price volatility, the
unidirectional link from the price volatility to the option
trading volume may break down.
Because informed traders may be more attracted to
the options markets than to the stock markets due to the
low transaction costs and higher leverage available in
the options markets, option trades may first reflect the
information on the future price volatility due to the fact
that option pricing formulas need this volatility to deter-
mine the option price (Easley et al., 1998). Thus, the option
trading volume may precede the future price volatility
if the option trades are largely initiated by informed
traders. However, the hedge-related uses of options suggest
that the future price volatility may precede the option trad-
ing volume because a higher future price volatility leads to
a greater use of options. So far, Sarwar (2003) appears to
be the only study that examines the direct relations between
the future price volatility and the option trading volume in
currency options markets for the British pound. His results
suggest that the volume of British pound options has
a strong predictive power for the future volatility of the
US dollar/British pound exchange rate. Whether such
results hold in other options markets remain to be seen,
however.
This study examines the informational role of the tradingvolume of equity index futures options in predicting thefuture price volatility of the underlying index futures.The sample options markets examined here and the pro-cedures used to analyse the volume-volatility relations inthese markets are distinctly different from those of Sarwar(2003). Specifically, the present sample options marketsdiffer from those studied by Sarwar in the followingrespects. First, unlike the thinly-traded market for theBritish pound options examined by Sarwar, this studyexamines the market for the S&P 500 index futures optionswhich is the largest equity futures options market on theChicago Mercantile Exchange in terms of daily tradingvolume. Second, jumps are less likely to occur in equityindices than in individual currencies because of the diversi-fication afforded by the indices. Third, Sarwar examinesonly the European currency options written on theBritish pound, while the S&P 500 futures options of thisstudy are all American-style options. Finally, because theS&P 500 futures options market allows traders a widerange of trading choices with respect to the expirationdates and strike prices, this multidimensional tradingchoice enhances the ability of informed traders to hidetheir trades in the market than is possible in the thinly-traded British pound options market. Hence, the volume–volatility relations in the S&P 500 futures options marketcould be very different from those uncovered by Sarwar(2003) in the British pound options market. Moreover,the study uses a pooled cross-sectional and time-seriesdata framework to adequately capture the subtle volume-volatility relations that could emerge from the multidimen-tional trading aspects of the S&P 500 futures options.
Besides examining the informational role of the futuresoptions volume for the future volatility of the underlyingfutures index, the study also investigates if the relationsbetween option trading and price volatility vary betweenthe rises and falls in the expected future price volatility.Examining such relations may uncover certain volatilitymovements that attract a greater trading activity fromthe informed and uninformed traders.
II . METHODOLOGY
This study uses variants of the causality testing approachesof Granger and Newbold (1977) and Granger (1969)to investigate the relations between price volatility andoption trading volume. Many researchers conductGranger causality tests by estimating a vector autoregres-sive (VAR) model and testing zero restrictions on thelagged parameters (see, for example, Chan and Chung,1993; Chatrath et al., 1995, 1996). The VAR model speci-fies that each endogenous variable depends upon its ownlagged values and the lagged values of the other endogen-ous variable (s) involved. To account for the possibility
1198 G. Sarwar
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
of contemporaneous interaction among the variables, onecan add the current value of one variable as an additionalexplanatory variable in the VAR equation of the othervariable (s) involved (Koch, 1993; Easley et al., 1998;Kyriacou and Sarno, 1999; Chan et al., 2002).
The VAR equations for analysing the dynamic relationsbetween the expected future price volatility and optionvolume series are specified as:
It ¼ 8 þXk
j¼1
91jIt�j
þXm
i¼0
81iVt�i þ et, t ¼ 0, 1, . . . , n ð1Þ
and
Vt ¼ 9 þXk
j¼0
92jIt�j
þXm
i¼1
82iVt�i þ kt, t ¼ 0, 1, . . . , n ð2Þ
where It denotes the future price volatility, Vt is the optiontrading volume, 8 and 9 denote the intercepts, and et and ktare the disturbance terms.
The series It and Vt are assumed to be stationary tocircumvent the problem of spurious regressions. If theseseries are nonstationary, they can be transformed tostationarity by differencing prior to parameter estimation(Griffiths et al., 1993).
Most investigators using aggregate data on optionsestimate parameters of Equations 1 and 2 either by theordinary least-square method or by a simultaneous equa-tion estimator (Chan and Chung, 1993; Koch, 1993;Chatrath et al., 1995, 1996; Kyriacou and Sarno, 1999).Similarly, researchers employing interday or intradaydata on stock options use the ordinary least-square methodto estimate variants of Equations 1 and 2 (Bhar andMalliaris, 1998; Easley et al., 1998; Chan et al., 2002).Such estimation approaches make no distinction betweenthe time series and cross-sectional properties of the optionsdata even though option observations are actually clus-tered in time and across cross-sections.
Previous empirical research on option pricing and actualdata on options trading strongly suggest the considerationof three estimation issues in Equations 1 and 2. First, vola-tility smiles studies provide ample evidence that impliedvolatilities of the underlying assets, which approximatethe expected future price volatility of the assets, vary bythe option’s moneyness (Heston, 1993; Bakshi et al., 1997;Das and Sundraram, 1999). Second, since price vola-tilities vary overtime as shown by several studies on thestochastic-volatility option pricing models (Melino andTurnbull, 1990; Heston, 1993; Nandi, 1996; Bakshi et al.,1997), the future price volatility expected at a given point-
in-time for, say, the next thirty days is likely to be differentfrom volatilities expected for other future time periods. Infact, Whaley (1982, 1986) and Dumas et al. (1998) empha-size the use of maturity-specific implied volatilities for pri-cing options. Thus, the future price volatility is expected tovary with the option’s time-to-maturity, which in turn sug-gests variation of the volatility–volume relations withrespect to the option’s time-to-maturity. In a related con-text, Nandi (1996) and Whaley (1982) recognize the impor-tance of the option’s time-to-maturity and moneynessclasses in explaining the option pricing biases, and thusthey estimate separate regressions for each of the severalsuch classes defined for their data. Third, actual optionsmarkets trade a series of options with different expirationdates and strike prices, allowing traders a wide range oftrading choices at any given time. This multidimensionaltrading choice enhances the ability of informed traders tohide their trades in the market, and thus it may blur theevidence of information-related trading in options if thevolatility–volume relations are analysed using data aggre-gated over the option’s moneyness and time-to-maturityclasses. These three considerations strongly suggest thatthe volume–volatility relations in options markets varyacross the option’s time-to-maturity and moneynessclasses. Thus, VAR Equations 1 and 2 need to be modifiedto account for such inter-class distinction of the volatility–volume relations.
This paper attempts to capture the cross-sectionaland time-series properties of the options data, as well asthe possible inter-class distinction of the volume-volatilityrelations, in a single pooled cross-sectional and time-seriesequation for each endogenous variable. To implement thisapproach, the time-series observations of options are firstordered by the option’s time-to-maturity classes, so that alloptions with a time-to-maturity of T months are stackedfirst, followed by options with a time-to-maturity of Tþ1months, and so on. These options are then divided intothree moneyness classes: near-the-money, in-the-money,and out-of-the-money. The product of the time-to-maturityclass and the moneyness class yields a cross-section whosedaily observations are tracked overtime until the option’sexpiration date. In this data scheme, the time-series obser-vations of cross-section one (near-the-money options withT maturity) are stacked on top, followed by observationsof cross-section two (in-the-money with T maturity) andcross-section three (out-of-the-money with T maturity),respectively. Similarly, cross-section four (near-the-moneywith Tþ1 maturity) is stacked under cross-section three,and this scheme is continued for other cross-sections.The resulting data show unequal number of observationsunder each cross-section, yielding an unbalanced pooledcross-sectional and time-series data structure.
This pooled cross-sectional and time-series data struc-ture is used to analyse the volatility–volume relationsin the sample options. The volume–volatility relations are
Informational role of option trading volume 1199
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
estimated using the error components and dummy variablemodels. To express these models succinctly, the VAREquations 1 and 2 are rewritten as:
Irt ¼ 8r þXK
J¼1
91jIrt�j
þXm
i¼0
81iVrt�i þ ert, t ¼ 0, 1, . . . , n ð3Þ
and
Vrt ¼ 9r þXk
j¼0
92jIrt�j
þXm
i¼1
82iVrt�i þ urt t ¼ 0, 1, . . . , n ð4Þ
where r refers to a cross-sectional unit (r¼ 1, 2, . . . ,N ), Irt�j
is the price volatility of cross-section r at time t�j, Vrt�i isthe option volume of cross-section r at time t�i, 8r (9r) isthe intercept of cross-section r in Equations 3 and 4, andert (urt) is the random error of cross-section r at time t inEquations 3 and 4. The random errors ert and urt areassumed to be independently and identically distributedwith a zero mean and constant variance.
The error components and dummy variable modelsassume that all behavioural differences between cross-sections are captured by the intercept. Thus, the modelsassume that each cross-section has its own, distinct inter-cept but all cross-sections share the same slope coefficients(see Griffiths et al., 1993 for details of these models). Thedummy variable model assumes that the intercepts arefixed parameters. However, the error components modelassumes that the intercepts are independent random vari-ables, so that we can write 8r¼8þ qr and 9r¼9þ 0r, where8 and 9 are unknown mean intercepts and qr and 0r areunobservable random disturbances that account for indi-vidual differences in cross-sections. Both models are usedfor analysis of the volume–volatility relations because thechoice between fixed and random effects is unclear a priorifrom the sample data.
If the volume and volatility series in Equations 3 and 4have contemporaneous interactions, then only 810 and 920are expected to be significant. However, if the linkages takesome time, the coefficients of the lagged price volatility oroption volume series would be significant. The null hypoth-esis, H0, that the future price volatility does not havea predictive power for the option trading volume isexpressed as:
H0 : 921 ¼ 922 ¼ � � � ¼ 92k ¼ 0 ð5Þ
that is, under H0 past values of the future price volatilitydo not have a predictive power for option volume. Thenull hypothesis is a joint hypothesis of zero-value lagged
coefficients that can be tested using an F-statistic for therestricted 921 ¼ 922 ¼ � � � ¼ 92k ¼ 0 and unrestricted equa-tion system (Griffiths et al., 1993; Koch, 1993). Similarly,the null hypothesis that the option volume does not have apredictive power for the future price volatility can be testedby simply replacing coefficients 92j in Equation 5 with 81i(i¼ 1, . . .,m). A rejection of the joint hypothesis of zero-value lagged 81i would be consistent with the information-related uses of options (Easley et al., 1998). Similarly, arejection of the joint hypothesis of zero-value lagged 92jis considered to be supportive of the hedge-related usesof options. If the null hypothesis in Equation 5 is rejected,the timing and direction of the predictive power of laggedvolatility terms can be examined by testing the significanceof individual lagged coefficients on the basis of a t-test(Easley et al., 1998). The number of lagged terms to includein Equations 3 and 4 is determined from the Schwartz andAkaike information criteria.
III . THE DATA
This study uses the daily closing transaction data forthe S&P 500 equity futures and futures options traded onthe Chicago Mercantile Exchange (CME) for the period2 January 2000 to 31 December 2000. The data areobtained from the Futures Industry Institute. The datalist for each futures option the opening, high, low, andclosing option prices; the date of the trade, maturity, andexercise price; and actual trading volume, open interest,and implied volatility. Similarly, the equity futures datalist for each contract the opening, high, low, settle, andclosing futures prices; the date of the trade and contractmaturity; and actual trading volume, delivery intent, andopen interest. The closing prices for the S&P 500 index(S&P) futures and futures options correspond to 3:15 PMcentral time, and both contracts are traded side by side onCME. The underlying futures for a futures option contractis the next expiring futures.
The daily data on futures options have a beginning sam-ple of 44,757 observations. Following Aı̈t-Sahalia and Lo(1998), options with time-to-maturity of less than a dayand with price less than $0.05 are dropped. In addition,options with a trading volume of less than two contractsare excluded. Aı̈t-Sahalia and Lo argue that prices ofoptions with a very low trading volume are notoriouslyunreliable. These filters yield a final sample of 37 883 obser-vations, which include 15 447 call observations and 22 436put observations.
The S&P futures options are American-style, and theunderlying asset is an index futures, the most likely casefor which jumps in prices are less likely to occur because ofthe diversification afforded by the index. Since the S&Pfutures options trade alongside the highly liquid futurescontracts, Daigler (1994) argue that the pricing of futures
1200 G. Sarwar
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
options will be more accurate than that of options on cash
assets because traders know the current price of the under-
lying futures.
The future price volatility of the S&P futures is approxi-
mated by the implied volatility using the procedures of
Whaley (1986) and Barone-Adesi and Whaley (1987).
Empirical evidence appears to favour an implied volatility
estimate over a historical volatility estimate as a predictor
of ex-post volatility. Chirac and Manaster (1978), Merville
and Pieptea (1989), Poteshman (2000) and Chernov (2001)
argue that implied volatilities should theoretically provide
the best forecasts of the expected future volatilities because
option prices can impound all publicly available informa-
tion. Their empirical results from the US equity index
options markets generally support the superiority of
the implied volatility as an estimate of expected future
volatility. Mayhew and Stivers (2003) show that for the
most actively traded CBOE stock options during 1988
and 1995, implied volatilities outperform GARCH volati-
lities and subsume all information in return shocks. They
also show that compared to GARCH volatilities, implied
volatilities of equity index options provide reliable incre-
mental information about the future firm-level volatility.1
In addition, Sarwar (2003) demonstrates that the overall
relations between the option volume and the expected
future price volatility in the British pound options markets
are unaffected whether the future price volatility is approxi-
mated by the implied volatility or by the GARCH volati-
lity.
This study uses Whaley’s (1986) approach to estimate
the maturity-specific implied volatility for each traded
option. These maturity-specific volatilities are updated
daily until the option’s expiration date.2 This approach to
volatility estimation is similar in spirit to that of Dumas
et al. (1998) in which they estimate maturity-specific
implied volatility for each traded option and then
revise volatility weekly. The continuously compounded
Treasury bill rate interpolated to match the maturity
of the option is used for the risk-free rate needed for
estimation of volatilities. The T-bill rates are the average
of the ask and bid discounts, and are hand-collected
from the Wall Street Journal. The estimation of implied
volatilities does not require the time series of dividends
on the S&P 500 index since the transaction prices of S&P
futures reflect these dividends.
The time-series data of implied volatilities and trading
volume is ordered by cross-sections according to
the pooled cross-sectional and time-series data structure
of section II. Since the data contain multiple options
for a cross-section defined with respect to the option’s
moneyness and time-to-maturity on many trading days
(for example, several in-the-money March 2000 options
traded on 3 January 2000), the average of maturity-specific
implied volatilities and the sum of the trading volume of
options are used to represent the daily volatility and
volume, respectively, for such cross-sections.3
IV. RESULTS
Table 1 presents the summary statistics of the sample data
on S&P futures options. The implied volatility of the index,
a proxy for the expected future price volatility, is very
unstable, varying from 2.14% to 93.85% during 2000.
The trading volume shows a wide range across the options,
with a mean value of 105 contracts and a standard devia-
tion of 279 contracts. The time to expiration of the options
averaged about two months for calls and two and a half
months for puts. The mean exercise value of the options is
at 94% of the mean spot value of index futures, but the
exercise value is nearly three times more volatile than the
spot value.
Table 2 presents the Phillips–Perron unit root tests for
examining the stationarity of the option volume and
implied volatility series. The hypothesis that the option
volume series is nonstationary is rejected at the 0.01
level.4 The implied volatility series is, however, nonstation-
ary at the 0.01 level. When first-differenced, the volatility
series is stationary. Therefore, the option volume and first-
differenced volatility (volatility change) series are used in
the subsequent analysis of volume – volatility relations to
circumvent the problem of spurious regressions.
1 In addition, unlike the implied volatilities in which each specific option maturity has its own distinct implied volatility, the GARCHvolatilities do not distinguish between maturity-specific volatilities and thus the same conditional volatility is used for the expected futurevolatility of all option maturities. Thus, GARCH procedures do not provide maturity-specific volatilities which are important inour pooled cross-sectional and time-series analysis of the volume-volatility relations.2Although the futures options data from the CME list implied volatility of the futures options, the CME does not specify the nature ofthese volatilities and the procedures used to compute them. Thus, implied volatilities are estimated using Whaley’s (1986) procedure.However, the mean and variance of the estimated implied volatilities are very close to those reported by the CME.3A futures option is considered to be near the money when the ratio of the spot index value to the exercise index value is between 0.975and 1.025. A call (put) option is out-of-the-money (in-the-money) when the spot index to exercise index ratio is less than 0.975, and a call(put) option is in-the-money (out-of-the-money) when the spot index to exercise index ratio is greater than 1.025. Slight variations inthese upper and lower limits of the moneyness classes do not materially affect the results in Tables 3–7.4 The results of the Phillips–Perron unit root tests are not affected when both trend and intercept terms are allowed or ignored in the tests.
Informational role of option trading volume 1201
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
Table 3 reports the regression results of the relation of
option volume and future price volatility from the error
components and dummy variable models. The table
also presents the results from the ordinary least-square
method, which assumes the same intercept and slope coef-
ficients across cross-sections, for comparative purposes.
The Akaike information and Schwartz criteria suggest
the use of five lags in the regressions. The partial autocor-
relation functions of the series also indicate that the partial
autocorrelation coefficients for both option volume and
price volatility are not significant beyond five lags.5
In Table 3, the results of Hausman test under the error
components model strongly suggest the existence of cross-
sectional random effects in intercepts. Similarly, the F-test
of the dummy variablemodel supports the presence of cross-
sectional fixed effects in intercepts.6 These results indicate
5 The results that follow are not materially affected when estimation is executed using an asymmetric lag structure that allows for higherthan five lagged terms in either the volatility or the volume series.6 The estimation of dummy variable model for Equations 3 and 4 also provides the intercept and t-value for each of the 40 distinct cross-sections in the data. These intercepts are not reported in Table 3 for brevity.
Table 1. Summary statistics for S&P 500 futures options data
Panel A: Aggregate datavariable Mean Standard deviation Minimum Maximum
Panel A: Aggregate dataOption price ($) 27.81 57.75 0.05 704.5Implied volatility (%) 24.21 7.71 2.14 93.85Days to expiration 75.20 64.34 2.00 335.00Exercise price (index points) 1388.09 180.86 850.0 2050.00Spot index value (index points) 1478.33 67.76 1279.60 1673.20Volume (option contracts) 105.32 279.62 2.0 7199.0
Panel B: Call optionsOption price ($) 31.18 67.38 0.05 653.00Implied volatility (%) 20.16 5.23 2.14 81.54Days to expiration 64.70 58.20 2.00 335.00Exercise price (index points) 1510.76 132.07 900.0 2050.0Spot index value (index points) 1479.24 68.01 1277.82 1672.03Volume (option contracts) 101.99 280.92 2.0 5805.00
Panel C: Put optionsOption price ($) 25.49 49.93 0.05 704.50Implied volatility (%) 27.17 7.83 2.14 93.85Days to expiration 78.06 68.00 2.00 335.00Exercise price (index points) 1303.64 160.40 850.00 2025.00Spot index value (index points) 1477.75 67.45 1283.82 1676.03Volume (option contracts) 107.61 278.69 2.0 7199.0
Notes: This table provides summary statistics for the daily closing prices of Chicago Mercantile Exchange call and put options on theS&P 500 index futures for the period 4 January 2000 to 31 December 2000. The aggregate sample has 37 883 observations, which include15 447 call and 22 436 put observations.
Table 2. Tests for stationarity of the volume and volatility series
Phillips – Perron unit root Z-test
Variable Aggregate sample Call options Put optionsVolume (level form) �38.56 �24.80 �26.03Implied volatility (level form) �2.99 �2.47 �2.02Implied volatility (first difference form) �28.19 �17.74 �19.25Critical Z-test value �6.25 �6.25 �6.25Total observations 37 883 15 447 22 436
Notes: This table shows the results of the Phillips – Perron unit root tests for the trading volume and implied volatilityseries of S&P 500 futures options from 4 January 2000 to 31 December 2000. The critical Z-test value for a sample ofinfinite observations came from Table 6 of Dickey and Fuller (1981, p. 1063).
1202 G. Sarwar
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
that the relations between option volume and expected
future price volatility vary across the time-to-maturity and
moneyness classes, and that the error components and
dummy variable models provide a reasonable framework
to capture these distinct cross-sectional effects. The results
undermine the use of ordinary least-square method and
other estimators that do not account for the differential
cross-sectional effects in volume–volatility relations.
The aggregate results in Table 3 do not show significant
contemporaneous feedbacks between price volatility and
option volume under the error components and dummy
variable models. Since the aggregate results average out
the contemporaneous effects of the call and put results,
the absence of contemporaneous effects in the aggregate
sample does not necessarily imply the absence of such
effects in the call and put samples. It is possible to get
insignificant contemporaneous feedbacks in the aggregate
sample if these feedbacks are positive for puts and negative
for calls, or vice versa. Interestingly, the least-square
method indicates significant and positive contemporaneous
interactions between volatility and option volume in the
aggregate sample. Hence, procedures that ignore the differ-
ential cross-sectional effects in volume–volatility relations
may provide misleading conclusions about the simulta-
neous effects between volatility and volume.
The lagged volatility terms jointly have a significant
positive effect on the option volume. The positive and
significant responses of volume to volatility at lags two
through five are consistent with the hedge-related uses of
S&P futures options. Thus, a rise in the expected future
volatility has a delayed persistent, positive influence over
the trading volume of futures options. Such a hedging role
Table 3. Option volume and implied volatility regressions: aggregate results
Dependent variable
Independent variableVolume (Vt) Implied volatility change (It)
EC Model DV Model OLS EC Model DV Model OLS
Intercept 234.4 �0.11 249.6 0.0001 0.01 �0.0006(4.62) (0.00) (11.75) (0.03) (2.16) (0.69)
It 568.2 346.0 1207.7 – – –(1.34) (0.74) (3.60)
It�1 471.2 618.8 177.68 �0.86 �0.85 �1.16(0.91) (1.17) (0.35) (83.73) (83.1) (90.5)
It�2 2765.2 2750.2 2936.3 �0.66 �0.66 �0.97(4.78) (4.74) (4.95) (45.94) (45.5) (48.9)
It�3 2787.5 2827.2 2624.9 �0.34 �0.33 �0.49(4.87) (4.95) (4.43) (21.11) (20.9) (21.7)
It�4 1902.7 1857.1 2110.3 �0.21 �0.21 �0.28(3.94) (3.85) (4.24) (15.65) (15.5) (14.5)
It�5 1441.6 1372.9 1673.7 �0.16 �0.15 �0.28(4.47) (4.26) (5.02) (17.07) (16.7) (21.46)
Vt – – – 0.00003 0.00003 0.00002(0.82) (0.74) (3.60)
Vt�1 0.008 0.018 �0.015 0.00003 0.00003 0.00004(0.61) (1.38) (1.13) (7.79) (7.71) (8.90)
Vt�2 0.12 0.12 0.13 �0.00007 �0.00007 �0.00007(9.23) (9.32) (9.61) (2.00) (1.93) (1.30)
Vt�3 0.31 0.29 0.35 �0.00001 �0.00001 �0.00002(24.4) (23.67) (28.08) (3.50) (3.43) (4.53)
Vt�4 0.13 0.13 0.14 �0.00002 �0.00001 �0.00002(10.34) (10.37) (11.02) (4.24) (4.17) (3.31)
Vt�5 0.03 0.03 0.04 �0.00003 �0.00002 �0.00001(2.73) (2.90) (2.68) (0.83) (0.72) (2.63)
R2 0.18 0.33 0.24 0.64 0.87 0.71F-statistic 15.04** 13.0** 27.0** 15.8** 15.4** 23.3**Hausman test 527.6** – – 3167.0** – –
(H0: No random effects)F-test – 7.7** – – 65.8** –(H0: No fixed effects)
Notes: The OLS, EC model, and DV model denote the ordinary least-square, error components model, and dummy variable model,respectively. Absolute t-values are reported in parentheses. The F-statistic tests the joint hypothesis that the five lagged coefficientsof Vt (It) are zero when the dependent variable is It (Vt). Two asterisks (**) indicate significance at the 0.01 level.
Informational role of option trading volume 1203
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
of futures options is both expected and reasonable. A find-
ing that did not support the hedging role of options would
be difficult to reconcile with the extensive use of options
markets for portfolio protection purposes. The positive
effects of lagged volatilities on the option volume may also
be related to the response structure of volume itself in
which the current volume displays a delayed but persistent
positive response to its lagged values over five trading days.
Like the lagged volatility terms, the lagged option
volumes jointly have a significant predictive power for
the future volatility of the S&P futures, a result consistent
with the information-related uses of S&P futures options.
However, the individual effects of lagged volumes on
volatility are mixed, with the volatility initially rising with
a rise in the previous-day volume and then subsequently
falling with rises in volumes on days two to five. These
volatility reactions to volume could be consistent with an
overreaction effect. Also, such reactions of volatility to
lagged volumes may be related to the response structure
of volatility itself in which the current volatility displays a
delayed but persistent negative response to its lagged values.
The individual effects of lagged volumes on volatility
also show subtle differences between the results of error
components and dummy variable models and those of
the least-square method. The least-square method indicates
that lag-two and lag-five volumes have statistically insig-
nificant and significant effects on volatility, respectively,
but the error components and dummy variable models
show exactly the opposite effects. These differential results
stem from the assumption of non-differential cross-
sectional effects under the least-square method.7 By con-
trast, the signs and statistical significance of all regression
coefficients, except the intercept, are similar between the
error components and dummy variable models. However,
the dummy variable model has a larger R2 value in both
regressions, and thus it may provide a better fit of the
volume–volatility relations than the error components
model.
The results are somewhat different when the aggregate
option volume is decomposed into the call and put
volumes. The main new result is that volatility and volume
have now significant contemporaneous feedbacks.
In Table 4, the put results show significant and positive
contemporaneous feedbacks between option volume and
future price volatility, suggesting that arbitrage and other
market forces operate efficiently to produce quick and
strong daily interactions between volume and volatility.
As in the aggregate results, the lagged volatilities jointly
have a significant positive effect on the put volume.
This positive influence suggests the lead of the expected
future price volatility over the put trading volume and is
consistent with the hedging-based uses of futures put
options. Also like the aggregate results, the lagged put
volumes jointly have a significant predictive power with
respect to the future price volatility, a result supportive
of the information-related uses of futures put options.
Hence, the trading volume of futures options is not a
non-informative outcome of the trading process, but it
actually contains valuable information about the future
price volatility. The lagged put volumes individually have
mixed effects on volatility, which are again similar to the
effects in the aggregate analysis. Further, the Hausman and
F tests continue to suggest the existence of significant
random and fixed effects in cross-sectional intercepts,
respectively.
The call results in Table 5 are nearly similar to the put
and aggregate results, with the exception of contempora-
neous relations between volatility and call volume. The call
results suggest significant and negative contemporaneous
feedbacks between volume and volatility, rather than
positive feedbacks as under the put options. The negative
feedbacks suggest that a rise in volatility quickly leads to
a fall in the call trading volume, and this fall in volume in
turn raises volatility. Such a negative impact of volatility
on volume may occur if the actual price volatility turned
out to be lower than the expected price volatility even
though the change in volatility is still positive.
However, similar to the put and aggregate results, the
call results show that lagged volatilities jointly have a sig-
nificant positive effect on the call volume. This positive
influence suggests the lead of the expected future price
volatility over the call trading volume and points to the
hedging role of futures call options. Also, like the aggregate
and put results, the lagged call volumes jointly have a sig-
nificant predictive ability for the expected future volatility,
a result that reinforces the information-related uses of S&P
futures options. In a related study, Easley et al. (1998) find
support for the information-based uses of stock options
when they examine the relations between the stock price
changes and the informationally defined option volume.
The results discussed so far implicitly assume that the
rises and falls in volatility have symmetric effects on the
option volume. It is probable that option traders are more
sensitive to increases in volatility (bad news) than they are
to decreases in volatility (good news). For instance, an
unexpected increase in volatility may increase the perceived
riskiness of equity positions, thus inducing portfolio
managers to seek additional risk protection from options.
But portfolio managers may not be willing to cut their
option positions at the first sign of a lower than expected
volatility because of the potential financial risks associated
with such a move. Tan (2001b) cites support for such asym-
7The equations for It and Vt are also estimated by the seemingly unrelated regression (SUR) method to check the possible impactof accounting for error-related bias on the estimated parameters. The SUR-based results are very similar to the OLS results in Table 3.
1204 G. Sarwar
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
metric effects of price volatility on option trading volume.
Nelson (1990), Schwert (1990) and Engle and Ng (1993)
suggest that good and bad news in returns have asymmetric
effects on the future price volatility. The present study
examines if good and bad news in the expected future
price volatility have asymmetric effects on the trading
volume of futures options.
Table 6 reports the asymmetric effects of positive and
negative changes in volatility on the option volume in the
aggregate sample.8 The results indicate that the latest
reduction in volatility individually as well as the previous
reductions in volatility jointly have significant downward
effects on the option volume. Similarly, the lagged increases
in volatility jointly have significant positive effects on
the option volume. These results clearly illustrate the role
of options markets as venues for hedging risks. Since
the lagged increases in volatility as well as the lagged
reductions in volatility have a significant predictive power
for the trading volume, it is not surprising to find that the
combined effect of the lagged rises and falls in volatility on
8The negative (positive) change in volatility variable in Table 6 is constructed as the product of the negative (positive) change in volatilityand a dummy variable that takes the value one when the volatility change is negative (positive) and zero otherwise.
Table 4. Option volume and implied volatility regressions: put options
Dependent variable
Volume (Vt) Implied volatility change (It)
IndependentVariable
ECmodel
DVmodel
ECmodel
DVmodel
Intercept 285.5 �137.4 �0.012 0.016(4.28) (0.50) (2.06) (2.90)
It 7425.3 7910.4 – –(9.90) (9.10)
It�1 5116.0 5692.8 �0.85 �0.85(5.39) (5.83) (63.84) (63.2)
It�2 6248.9 6686.2 �0.58 �0.57(6.06) (6.44) (29.83) (29.57)
It�3 4557.6 4724.9 �0.29 �0.28(4.50) (4.67) (13.95) (13.82)
It�4 1870.0 2126.3 �0.16 �0.16(2.20) (2.51) (8.92) (8.86)
It�5 999.4 983.0 �0.08 �0.08(1.92) (1.88) (7.77) (7.68)
Vt – – 0.00004 0.00004(9.20) (9.10)
Vt�1 0.020 0.026 0.00005 0.00005(1.08) (1.35) (13.9) (13.8)
Vt�2 0.11 0.11 �0.00002 �0.00002(6.05) (6.03) (4.31) (4.17)
Vt�3 0.25 0.25 �0.00001 �0.00001(13.8) (13.6) (2.83) (2.78)
Vt�4 0.16 0.16 0.00007 0.00001(8.91) (8.86) (0.19) (0.27)
Vt�5 0.07 0.08 �0.00002 �0.00001(3.95) (4.17) (2.88) (2.79)
R2 0.22 0.35 0.68 0.90F-statistic 8.39** 8.97** 44.4** 44.04**Hausman test 44.5** – 44.6** –(H0: No random effects)F-test – 5.95** – 102.06**(H0: No fixed effects)
Notes: The EC model and DV model denote the error components model and dummy variable model, respectively.Absolute t-values are reported in parentheses. The F-statistic tests the joint hypothesis that the five laggedcoefficients of Vt(It) are zero when the dependent variable is It(Vt) Two asterisks (**) indicate significance at the0.01 level.
Informational role of option trading volume 1205
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
option volume is also positive. More importantly, however,is the finding that the rises and falls in volatility do nothave symmetric influences on the option volume. Rather,increases in lagged volatilities jointly exert a larger positiveeffect on option volume than do decreases in laggedvolatilities. Thus, option traders appear to be more sensi-tive to the rises in the expected future volatility (bad news)than to the falls in the expected future volatility (goodnews) of similar magnitude.
The asymmetric call and put results in Table 7 are simi-lar, and they corroborate the aggregate asymmetric resultsin Table 6. In both the call and put options, the jointpositive effects of lagged increases in volatilities on theoption volume are larger than those of lagged reductionsin volatilities. The only notable exception in the call andput results is the effect of the latest rise and fall in volatility
on the option trading volume. Under the put options, thecurrent rise and fall in volatility have the expected positiveand significant effects on the option volume. However,under the call options, the latest increase in volatilityalone has a significant effect on the option volume, butthis effect is unexpectedly negative. Such mixed effects incalls are perhaps not too unexpected, and they may resultfrom the difference between the expected and actual pricevolatilities and the subsequent adjustments in volatilityexpectations.
Overall, the results overwhelmingly support the conten-tion that the markets for S&P futures options are venuesfor information-related trading. The results also clearlyillustrate the role of markets for S&P futures options asvenues for hedging risks. In hedging their risks with the useof futures options, portfolio and risk managers appear to be
Table 5. Option volume and implied volatility regressions: call options
Dependent variable
Volume (Vt) Implied volatility change (It)
Independent variable EC model DV model EC model DV model
Intercept 334.8 2073.5 0.017 0.13(4.68) (5.24) (2.68) (8.92)
It �3626.8 �3945.4 – –(7.31) (7.44)
It�1 �2091.8 �1923.1 �0.81 �0.80(3.48) (3.13) (49.84) (49.36)
It�2 48.81 �244.1 �0.70 �0.69(0.07) (0.37) (34.15) (33.9)
It�3 2247.5 2313.3 �0.34 �0.33(3.39) (3.48) (14.29) (14.1)
It�4 1251.6 1127.9 �0.26 �0.26(2.23) (2.02) (13.27) (13.24)
It�5 1249.1 1352.6 �0.18 �0.18(3.11) (3.37) (12.78) (12.43)
Vt – – �0.00005 �0.00005(7.46) (7.44)
Vt�1 �0.038 �0.027 0.00006 0.00004(1.95) (1.41) (0.80) (0.62)
Vt�2 0.14 0.13 �0.00002 �0.00002(7.23) (6.66) (2.28) (2.35)
Vt�3 0.35 0.34 �0.00003 �0.00003(19.4) (18.7) (0.35) (0.40)
Vt�4 0.07 0.06 �0.00004 �0.00004(3.85) (3.45) (5.73) (5.81)
Vt�5 �0.003 �0.003 �0.00007 �0.00006(0.19) (0.14) (1.04) (0.94)
R2 0.21 0.36 0.66 0.86F-statistic 16.12** 15.9** 9.83** 10.15**Hausman test 63.9** – 75.97** –(H0: No random effects)F-test – 8.09** – 45.8**(H0: No fixed effects)
Notes: The EC model and DV model denote the error components model and dummy variable model, respectively.Absolute t-values are reported in parentheses. The F-statistic tests the joint hypothesis that the five lagged coefficients ofVt(It) are zero when the dependent variable is It(Vt) Two asterisks (**) indicate significance at the 0.01 level.
1206 G. Sarwar
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
more responsive to the rises in the expected future volatilitythan they are to the falls in the expected future volatility.Compared to Sarwar (2003) who reports that the informa-tion-related trading explains more of the trading volume ofBritish pound options than hedging, the present study findsthat both the hedge-related and information-related trad-ing are significant forces in explaining most of the tradingvolume of S&P futures options.
V. CONCLUSIONS
This paper examines the dynamic relations between theexpected future price volatility of the S&P 500 indexfutures and the trading volume of S&P 500 futures optionsduring 2000 to explore the informational role of the optionvolume in predicting the future price volatility of the indexfutures. The future price volatility is approximated by theimplied volatility of the S&P 500 futures options. Using a
pooled cross-sectional and time-series data framework, thepaper uses the error components and dummy variablemodels to allow for the relations between volatility andvolume to vary by the option’s moneyness and time-to-maturity.
The results indicate that the relations between the optionvolume and the future price volatility vary across cross-sections defined with respect to the option’s time-to-matur-ity and moneyness. The error components and dummyvariable models provide a reasonable framework tocapture such cross-sectional effects in volume-volatilityrelations. These results undermine the use of estimatorsthat do not account for the differential cross-sectionaleffects in volume-volatility relations.
This study uncovers the presence of strong contempora-neous feedbacks between the price volatility and the trad-ing volume of call and put options, suggesting thatarbitrage forces operate efficiently to produce quick andstrong interactions between volume and volatility.
Table 6. Asymmetric effects of volatility on option volume: aggregate sample
Dependent variable: Volume (Vt)
Independent variable Error components model Dummy variable model
Intercept 40.28 (0.71) �152.7 (0.59)It(þ) 655.03 (1.24) 1095.3 (1.89)It(�) 3510.4 (4.08) 3018.6 (3.39)It�1(þ) 3206.1 (3.58) 3130.7 (3.43)It�1(�) 4057.7 (4.81) 4352.9 (5.14)It�2(þ) 7822.0 (8.54) 7958.5 (8.67)It�2(�) 2821.3 (3.17) 2684.7 (3.03)It�3(þ) 4748.3 (4.99) 4720.5 (4.97)It�3(�) 3996.6 (4.72) 3963.2 (4.70)It�4(þ) 4983.6 (5.51) 4798.3 (5.32)It�4(�) 2017.9 (2.60) 2103.1 (2.70)It�5(þ) 3236.1 (3.86) 3361.1 (4.02)It�5(�) 701.1 (1.80) 566.9 (1.45)Vt�1 �0.005 (0.43) 0.003(0.24)Vt�2 0.113 (8.54) 0.114(8.64)Vt�3 0.28 (22.48) 0.27(21.58)Vt�4 0.12 (9.07) 0.12 (9.05)Vt�5 0.02 (1.89) 0.03 (2.06)R2 0.20 0.35F-statistic (þI1�5¼0) 18.25** 18.59**F-statistic (�I1�5¼0) 7.20** 7.67**F-statistic (þI1�5¼�I1�5) 20.39** 20.40**F-statistic (all I1�5¼0) 18.38** 17.43**Hausman test 61.26** –(H0: No random effects)F-test (H0: No fixed effects) – 7.67**
Notes: This table reports the results of causality regressions when positive and negative changes in implied volatility aretreated as separate explanatory variables for the option volume. Positive (negative) changes in volatility at lag k aredenoted as kþ (k�) Absolute t-values are reported in parentheses. The results of the regressions when the dependentvariable is volatility change (It) are omitted for brevity as they are similar to the results in Table 3. The F-statistic (þI1�5)and F-statistic (�I1�5), respectively, test the joint hypothesis that the coefficients of positive and negative changes involatility at lags 1–5 are zero. Similarly, the F-statistic (þI1�5¼�I1�5) tests the joint hypothesis that the coefficients ofpositive and negative changes in volatility at lags 1–5 are equal to each other, and the F-statistic (all I1�5) tests the jointhypothesis that the coefficients for both the positive and negative changes in volatility at lags 1–5 are zero. Two asterisks(**) indicate significance at the 0.01 level.
Informational role of option trading volume 1207
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
The main result of the study is that the trading volume of
S&P 500 futures options has a strong predictive ability with
respect to the future price volatility of the S&P 500 index
futures, and that the markets for S&P 500 futures options
are venues for information-related trading. The study also
finds strong evidence that clearly illustrates the role of mar-
kets for S&P futures options as venues for hedging risks.
However, the rises and falls in the expected future price
volatility have asymmetric influences on the hedge-related
uses of futures options. In hedging their risk exposures,
portfolio and risk managers appear to be more responsive
to the rises in future price volatility than to the falls in
future price volatility.
The present results based on the interday data of index
options may have implications for the relations between
option volume and future price volatility in other options
markets. Because interactions between trading volume and
price volatility are likely to be stronger in intraday data of
index options and actively traded stock options than ininterday data of index options, the volume–volatilityresults of this study may also hold for such options data.This remains to be seen in future research, however.
REFERENCES
Aı̈t-Sahalia, Y. and Lo, A. W. (1998) Nonparametric estimationof state-price densities implicit in financial asset prices,Journal of Finance, 53, 499–548.
Anthony, J. H. (1988) The interrelation of stock and option mar-ket trading-volume data, Journal of Finance, 43, 949–61.
Antoniou, A. and Holmes, P. (1995) Futures trading, informationand spot price volatility: evidence for the FTSE-100 stockindex futures contract using GARCH, Journal of Bankingand Finance, 19, 117–29.
Back, K. (1993) Asymmetric information and options, Review ofFinancial Studies, 6, 435–72.
Bakshi, G., Cao, C. and Chen, Z. (1997) Empirical performanceof alternative option pricing models, Journal of Finance, 52,2003–49.
Table 7. Asymmetric effects of volatility on option volume: call and put options
Dependent variable: Volume (Vt)
Call options Put options
Error components Dummy variable Error components Dummy variableIndependent variable model model model model
Intercept 219.1 (2.88) 1839.3 (4.64) �11.1 (0.12) �297.5 (1.09)It (þ) �3577 (5.90) �3617.4 (5.49) 6164 (5.90) 7434.8 (6.51)It (�) �2066 (1.65) �2268.9 (1.77) 8045 (6.76) 7853.7 (6.09)It�1 (þ) �954 (0.73) �663.2 (0.50) 5202 (3.69) 5834.4 (4.09)It�1 (�) 3601 (2.76) 4053.2 (3.11) 6875 (5.83) 7136.2 (5.89)It�2 (þ) 5993 (4.40) 6100.6 (4.49) 1129.5 (7.94) 11320 (7.92)It�2 (�) 2251 (1.69) 1717.7 (1.29) 3327 (2.59) 3950.3 (3.07)It�3 (þ) 5683 (4.16) 5603.7 (4.00) 4844 (3.37) 4937.2 (3.42)It�3 (�) 3709 (2.86) 3886.1 (3.01) 5560 (4.52) 5564.4 (4.52)It�4 (þ) 3966 (2.87) 3850.0 (2.80) 4821 (3.65) 4803.1 (3.63)It�4 (�) 3108 (2.61) 3300.3 (2.78) 1874 (1.65) 2074.2 (1.83)It�5 (þ) 4668 (3.60) 5112.0 (3.95) 3676 (3.17) 3513.6 (3.03)It�5 (�) 365.8 (0.82) 436.6 (0.97) 110.2 (0.15) 83.8 (0.11)Vt�1 �0.04 (2.39) �0.03 (1.78) 0.02 (1.06) 0.02 (1.21)Vt�2 �0.12 (6.51) 0.12 (5.95) 0.10 (5.48) 0.11 (5.44)Vt�3 0.33 (18.16) 0.32 (17.59) 0.21 (11.32) 0.21 (11.28)Vt�4 0.06 (3.29) 0.03 (2.94) 0.15 (7.90) 0.15 (7.90)Vt�5 �0.01 (0.51) �0.007 (0.41) 0.05 (2.87) 0.06 (3.14)R2 0.23 0.38 0.24 0.36F-statistic (þI1�5¼ 0) 8.85** 9.16** 15.26** 15.07**F-statistic (�I1�5¼ 0) 3.31** 4.03** 8.64** 8.67**F-statistic (þI1�5¼�I1�5) 9.60** 10.30** 9.60** 8.06**F-statistic (all I1�5¼ 0) 13.70** 14.07** 9.01** 8.50**Hausman test 136.3** – 71.99** –(No random effects)F-test (No fixed effects) – 8.53** – 5.08**
Notes: This table reports the results of causality regressions when positive and negative changes in implied volatility are treated asseparate explanatory variables for the option volume. Positive (negative) changes in volatility at lag k are denoted as kþ (k�) Absolutet-values are reported in parentheses. The F-statistic (þI1�5) and F-statistic (�I1�5), respectively, test the joint hypothesis that thecoefficients of positive and negative changes in volatility at lags 1–5 are zero. Similarly, the F-statistic (þI1�5¼�I1�5) tests the jointhypothesis that the coefficients of positive and negative changes in volatility at lags 1–5 are equal to each other, and the F-statistic(all I1�5) tests the joint hypothesis that the coefficients for both the positive and negative changes in volatility at lags 1–5 are zero.Two asterisks (**) indicate significance at the 0.01 level.
1208 G. Sarwar
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
Barone-Adesi, G. and Whaley, R. E. (1987) Efficient analyticapproximation of American option values, Journal ofFinance, 42, 301–20.
Bhar, R. and Malliaris, A. G. (1998) Volume and volatility inforeign currency futures markets, Reveiw of QuantitativeFinance and Accounting, 10, 285–302.
Bhattacharya, M. (1987) Price changes of related securities:the case of call options and stock, Journal of Financial andQuantitative Analysis, 22, 1–15.
Black, F. (1975) Fact and fantasy in the use of options, FinancialAnalysts Journal, 31, 36–41.
Black, F. and Scholes, M. (1973) The pricing of optionsand corporate liabilities, Journal of Political Economy, 81,637–59.
Chan, K. and Chung, P. Y. (1993) Intraday relationships amongindex arbitrage, spot and futures price volatility, and spotmarket volume: a transaction data set, Journal of Bankingand Finance, 17, 663–87.
Chan, K., Chung, Y. P. and Chung, W.-M. (2002) The informa-tional role of stock and option volume, The Review ofFinancial Studies, 15, 1049–76.
Chan, K., Chung, P. Y. and Johnson, H. (1993) Why optionprices lag stock prices: a trading-based explanation, Journalof Finance, 48, 1957–68.
Chatrath, A., Ramchander, S. and Song, F. (1995) Does optionstrading lead to greater cash market volatility?, The Journal ofFutures Markets, 15, 785–803.
Chatrath, A., Ramchander, S. and Song, F. (1996) The role offutures trading activity in exchange rate volatility, TheJournal of Futures Markets, 16, 561–84.
Cherian, J. (1993) Option pricing, self-fulfilling prophecies,implied volatilities, and strategic interaction, Ph.D. disserta-tion, Cornell University.
Chernov, M. ( 2001) Implied volatilities as forecasts of futurevolatility, time-varying risk premia, and return variability,Working Paper, Division of Finance and Economics,Columbia Business School, New York, USA.
Chirac, D. P. and Manaster, S. (1978) The information content ofmarket prices and a test of market efficiency, Journalof Financial Economics, 6, 213–34.
Diagler, R. T. (1994) Financial Futures and Options Markets:Concepts and Strategies, HarperCollins College Publishers,New York.
Das, S. R. and Sundraram, R. K. (1999) Of smiles and smirks: aterm structure perspective, Journal of Financial andQuantitative Analysis, 34, 211–40.
Dumas, B., Fleming, J. and Whaley, R. E. (1998) Impliedvolatility functions: empirical tests, Journal of Finance, 53,2059–106.
Dickey, D. A. and Fuller, W. A. (1981) Likelihood ratio statisticsfor autoregressive time-series with a unit root, Econometrica,49, 1057–72.
Easley, D., O’Hara, M., and Srinivas, P. S. (1998) Option volumeand stock prices: evidence on where informed traders trade,Journal of Finance, 53, 431–65.
Engle, R. F. and Ng, V. K. (1993) Measuring and testing of theimpact of news on volatility, The Journal of Finance, 48,1749–78.
Fleming, J., Barbara, O. and Whaley, R. E. (1996) Tradingcosts and the relative rates of price discovery in stock,futures, and option markets, Journal of Futures Markets,16, 353–87.
Granger, C. W. J. (1969) Investigating causal relations byeconometric models and cross-spectral models,Econometrica, 37, 424–38.
Granger, C. W. J. and Newbold, P. (1977) Forecasting EconomicTime Series, Academic Press, New York.
Griffiths, W. E., Hill, R. C. and Judge, G. G. (1993)Learning and Practicing Econometrics, John Wiley & Sons,New York.
Heston, S. (1993) A closed-form solution for options withstochastic volatility with applications to bond and currencyoptions, The Review of Financial Studies, 6, 327–43.
Koch, P. D. (1993) Reexamining intraday simultaneity in stockindex futures markets, Journal of Banking and Finance, 17,1191–205.
Kyriacou, K. and Sarno, L. (1999) The temporal relationshipbetween derivatives trading and spot market volatility inthe UK: empirical analysis and Monte Carlo evidence, TheJournal of Futures Markets, 19, 245–70.
Llorence, G., Michaely, R., Saar, G. and Wang, J. (2002)Dynamic volume-return relation of individual stocks, TheReview of Financial Studies, 15, 1005–48.
Manaster, S. and Rendleman, R. J. (1982) Option prices aspredictors of equilibrium stock prices, Journal of Finance,37, 1043–57.
Mayhew, S. and Stivers, C. (2003) Stock return dynamics, optionvolume, and the information content of implied volatility,Journal of Futures Markets, 23, 615–46.
Mayhew, S., Sarin, A. and Shastri, K. (1995) The allocation ofinformed trading across related markets: an analysis of theimpact of changes in equity-market margin requirements,Journal of Finance, 50, 1635–54.
Melino, A. and Turnbull, S. M. (1990) Pricing foreign currencyoptions with stochastic volatility, Journal of Econometrics,45, 239–65.
Merville, L. J. and Pieptea, D. R. (1989) Stock price volatility,mean-reverting diffusion, and noise, Journal of FinancialEconomics, 24, 193–214.
Nandi, S. (1996) Pricing and hedging index options under sto-chastic volatility: an empirical examination, Working Paper96-9, Federal Reserve Bank of Atlanta, USA.
Nelson, D. (1990) Conditional heteroskedasticity in asset returns:a new approach, Econometrica, 59, 347–70.
Poteshman, A. M. (2000) Forecasting future volatility fromoption prices, Working Paper, Department of Finance,University of Illinois at Urbana-Champaign, IL, USA.
Sarwar, G. (2003) The interrelation of price volatility and tradingvolume of currency options, Journal of Futures Markets, 23,681–700.
Schwert, G. W. (1990) Stock volatility and the crash of 87, TheReview of Financial Studies, 3, 77–102.
Sears, S. M. (2000a) Equity options trading takes defensive turnas players react to stock market declines,Wall Street Journal,19 September, C16.
Sears, S. M. (2000b) Cisco, Dell and Ciena attract major activity,as traders retake positions in tech sector, Wall Street Journal,26 September, C19.
Sheikh, A. M. and Ronn, E. I. (1994) A characterization of thedaily and intraday behavior of returns on options, Journalof Finance, 49, 557–80.
Stephen, J. A. and Whaley, R.E. (1990) Intraday price change andtrading volume relations in the stock and stock optionsmarket, Journal of Finance, 5, 191–220.
Tan, K. (2001a) High put-call ratio shows defensive postures stilldominate, despite slip in volatility, Wall Street Journal, 19September, C14.
Tan, K. (2001b) Rate cut nudges volatility down, but investorsremain anxious, continue defensive hedging, Wall StreetJournal, 3 October, C11.
Informational role of option trading volume 1209
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14
Vijh, A. M. (1988) Potential biases from using only trades pricesof related securities on different exchanges, Journal ofFinance, 43, 1049–55.
Vijh, A. M. (1990) Liquidity of the CBOE equity options, Journalof Finance, 45, 1157–79.
Whaley, R. (1982) Valuation of American call options ondividend-paying stocks, Journal of Financial Economics, 10,29–57.
Whaley, R. (1986) Valuation of American futures options: theoryand empirical evidence, Journal of Finance, 41, 127–50.
1210 G. Sarwar
Dow
nloa
ded
by [
Flor
ida
Inte
rnat
iona
l Uni
vers
ity]
at 0
5:45
20
Dec
embe
r 20
14