CFA Institute

Implied Volatility
Author(s): Stewart Mayhew
Source: Financial Analysts Journal, Vol. 51, No. 4 (Jul. - Aug., 1995), pp. 8-20
Published by: CFA Institute
Stable URL: http://www.jstor.org/stable/4479853 .
Accessed: 25/10/2013 01:27

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

CFA Institute is collaborating with JSTOR to digitize, preserve and extend access to Financial Analysts
Journal.

http://www.jstor.org

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
 LITERATURE
REVIEW

Implied Volatility
StewartMayhew

This literaturereviewsummarizesthe academicresearchon option-impliedvolatility. It

describesalgorithmsfor calculatingimpliedvolatilityand variousweightingschemesusedto
derivea singlevolatilityestimatefromthepricesof multipleoptions,summarizesevidencein
the debateon whetherto use historicaldataor impliedvolatilityin forecasting,and reviews
severalotherpaperson the varioususes of impliedvolatility,includingmarketefficiency
studiesand event studies. This reviewalso suggeststhat impliedvolatilityis beingwidely
misusedin practice,describesthe schizophrenic notionof the volatilitysmile, and reviews
variousmethodsfor calculatinga risk-neutral densityfunctionconsistentwith optionprices
and the newgenerationof optionpricingmodels(suchas Rubenstein'simpliedbinomialtree
method)basedon theseimpliedvolatilities.

Qptionpricing formulas,such as the Black- deterministically over time, implied volatility is

Scholes formula, relate the price of an option interpreted to be the market's assessment of the
to the underlying asset price, the volatility of the average volatility over the remaining life of the
underlying asset, and other parametersthat influ- option.1 Option pricing formulas other than the
ence option prices. The underlying stock price and Black-Scholes or binomial also may be used to
the other parameters, including the strike price of calculateimplied volatilities. If the volatility of the
the option, time to expiration, interest rate, and underlying asset is itself a random process, as is
dividend yield of the underlying asset, are rela- assumed in "stochastic volatility" models, the
tively easy to observe. Given that these values are market prices of options can still be used to esti-
known, the pricingformularelates the option price mate the parametersof the underlying asset pro-
to the volatility of the underlying asset. Historical cess.2
stock price data may be used to estimate the Although the concept of implied volatility is
volatility parameter, which then can be plugged commonly associated with standard stock options
into the option pricing formula to derive option or stock index options, it is also quite useful in
values. As an alternative, one may observe the other contexts. Impliedvolatilitymay be calculated
market price of the option, then invert the option from the prices of exotic options, as demonstrated
pricing formula to determine the volatility implied by Ball, Torous, and Tschoegl (1985). Several au-
by the market price. The market's assessment of thors have examined implied volatility using op-
the underlying asset's volatility, as reflected in the tions on commodity or currencyfutures,3 and the
option price, is known as the impliedvolatilityof the prices of bond options can be used to estimate the
option. parametersof an underlying term structuremod-
Traditionally,implied volatility has been cal- el.4
culated using either the Black-Scholes formula or Option pricing formulas often cannot be in-
the Cox-Ross-Rubinstein binomial model. Under verted analytically, so implied volatility must be
the strictassumptions of the Black-Scholesmodel, calculatednumerically.In general, this calculation
implied volatility is interpreted as the market's is accomplished by feeding the value-price differ-
estimate of the constant volatilityparameter.If the ence,
underlying asset's volatility is allowed to vary
- CM,
C(oC) (1)

StewartMayhewis a doctoralstudentin financeat the Universityof into a root-finding program, where C( ) is an

at Berkeley.
California option pricing formula, c- is the volatility parame-

8 FinancialAnalysts Journal / July-August 1995

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
ter,5 and CMis the observed market price of the
option. Various algorithmscan be used to find the =-
U i (2)
value of o-that makes this expression equal to zero.
Choosing among them involves a tradeoffbetween One concern with using equal weights is that
robustness and speed of convergence. A simple the model is not, in fact, correct. The Black-
approach that is very slow but reliable is to try a Scholes model prices some options more accu-
series of differentvalues for o-and choose the one rately than others, and placing more weight on
that comes closest to satisfying condition (1). observations for which the model performs better
Sometimes known as the "shotgun method," this is reasonable. Trippi and Schmalensee and Trippi
approachis easy to implement but inefficientcom- dealt with this problem by simply throwing out
pared with other techniques such as the bisection options that are near expiration or far from the
method that for all practicalpurposes are just as money.
robust.6 Faster convergence can be achieved if an Another problem with equal weights is that
analytic expression is known for the option's some options are more sensitive to volatility than
"vega"-the derivative of the option price with others;estimationerrors(such as those induced by
respect to the volatility parameter.Such is the case price discreteness or asynchronous data) are likely
for the Black-Scholes formula, for which a New- to be higher for options whose prices are insensi-
ton-Raphson algorithm can usually achieve rea- tive to volatility. Therefore, placing more weight
sonably accurate estimates within two or three on options with higher vegas (higher sensitivities
iterations.7 Resorting to numerical procedures is to volatility) appears to be preferable to equal
not always necessary;for the special case of at-the- weighting. Latane and Rendleman (1976) sug-
money options, Brenner and Subrahmanyam gested this weighting scheme:
(1988)showed that the Black-Scholesformula can
be inverted to derive a simple formula for implied
volatility. 0-= / E i, (3)
N i=l

SWi
i=l

VOLAnILmES where the weights, wi, are the Black-Scholesvegas

IMPLIED
WEIGHTED-AVERAGE
Often, many options, which vary in strike price of the options. This forecast has the advantage of
and time to expiration, are written on the same weighting options according to their sensitivities,
but it is subject to the criticism that it is biased
underlying asset. If the Black-Scholesmodel held
exactly, these options would be priced so that they because the weights do not sum to 1. Chiras and
all have exactly the same implied volatility, which Manaster(1978)suggested weighting not by vegas
of course, is not the case. Systematic deviations but by volatility elasticities:
from the predictions of the Black-Scholes model,
often called the "volatilitysmile," are discussed in 8(7i Ci

a later section of this review. Even if market

participants were to price options according to 0-= (4)
N
5CO'
Black-Scholes, price discreteness, transactions
costs, and nonsynchronous trading would cause l=1oa.i Ci
observed implied volatilities to differ across op- Beckers (1981) and Whaley (1982) suggested
tions. choosing to minimize
In response to this problem, an earlybranchof
literaturesuggested calculatingimplied volatilities N

for each option and then using a weighted average wi[Ci - BSi()], (5)
of these implied volatilities as a point estimate of i=l

future volatility. The idea behind this approach is where Ci is the market price and BSi the Black-
simple: If the model is correct, then deviations Scholes price of option i. The weights, wi, may be
from the predicted prices represent noise, and chosen in many ways, the most obvious choices
noise can be reduced by using more observations. being equal weights or Black-Scholesvegas.
The simplest weighting scheme, used by Trippi Which of these many methods is best at pre-
(1977) and by Schmalensee and Trippi (1978), dicting volatility? Beckers (1981) addressed this
places equal weights on all N implied volatilities: question empirically using daily prices of equity

Financial Analysts Journal / July-August 1995 9

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
options from October13, 1975, to January23, 1976. Scott and Tucker (1989)examined the relative
He compared the forecasting ability of three mea- performance of various weighting schemes for
sures of implied volatility. The first was the mea- calculating implied volatility from currency op-
sure suggested by Latane and Rendleman and tions. Using transactions data for American-style
given in equation (3). The second was the qua- FX calls on the British pound, Canadian dollar,
draticloss function given by equation (5), and the deutschemark, yen, and Swiss franc from March
third was simply the implied volatility of the 14, 1983, to March13, 1987, these authors inverted
option with the highest vega.8 Beckersfound that the Garman-Kohlhagen (1983) currency option
the squared-error-minimizingtechnique led to bet- formula to calculate implied volatilities.10 They
ter forecasts than the Lataneand Rendlemanmea- examined three weighting schemes: vega weights,
sure. He also found, however, that using the minimized squared pricing error, and the method
implied volatility of the option with the highest that places all the weight on the option nearest the
vega outperforms either of the other two tech- money. They found that all three weights perform
niques. equally well and that adding historical volatility
Beckers'sresults seem to make the weighted- does not improve predictive accuracy. Additional
average approachto estimating implied volatilities research on the forecasting power of the implied
obsolete, because using the implied volatility of volatilityof currencyoptions has been reportedby
the nearest-in-the-moneycall option appears to do Fung, Lie, and Moreno (1990) and by Edey and
as well at forecasting future volatility as does any Eliot (1992). Turvey (1990) tested alternative
weighted average; it also has the advantage of weighting schemes for calculatingimplied volatil-
being easier to calculate. Nevertheless, a fair ities for options on soybean and live cattle futures.
amount of subsequent research has been devoted Maloney and Rogalski (1989) found that op-
to this field. Brenner and Galai (1981) found that tion prices reflect predictableseasonal patterns in
additional forecasting power can be achieved by volatility. Morse (1991) also looked at the season-
calculatingthe weighted-averageimplied volatility ality of implied volatility, finding that the differ-
several times during the day and averaging the ence between call and put implied volatility tends
results instead of using closing prices. This finding to drop on Fridaysand rise on Mondays. Resnick,
suggests that intraday frictions may significantly Sheikh, and Song (1993)described an "expiration-
affect implied volatility estimates.9 specific" weighted-average implied volatility, tak-
Whaley (1982) used a minimized-squared- ing into account monthly patterns that differ sys-
pricing-errorimplied volatility to investigate em- tematicallywith market capitalization.Franksand
pirically the various models for dividend-paying Schwartz (1991), using options on the Financial
Americancall options. Using weekly closing prices Times Stock Exchange Index from May 1984 to
for Chicago Board Options Exchange (CBOE)op- December 1989, examined the time-series proper-
tions on 91 dividend-paying stocks from January ties and macroeconomic determinants of Chiras-
1975 to February1978, he found that this measure Manaster weighted-average implied volatility.
of implied volatility is more accuratethan equally They found that shocks to implied volatilitydo not
weighted, vega-weighted, or elasticity-weighted persist for long and that leverage, inflation, and
average implied volatility. He also found that the long-term nominal interest rates help explain im-
different dividend models give approximatelythe plied volatility.
same implied volatilities. Weighted-average implied volatilities have
Gemmill (1986) looked at 13 equity options also been used to construct volatility indexes.
traded on the London Traded Options Market. Before the advent of index options, Gastineau
Using monthly closing prices from May 1978 to (1977) suggested a volatility index constructed
July 1983, he compared six different implied vola- from the implied volatilities of individual stock
tility weighting schemes-equal weights, elasticity options.11Recently, the CBOEintroduced its Mar-
weights, minimized squared pricing errors, near- ket VolatilityIndex (VIX).This index was designed
est the money only, farthest out of the money to represent the implied volatility of an at-the-
only, and farthest in the money only-to see money option on the S&P100 (OEX)Index with 22
which best predicts future volatility. Using a re- trading days to maturity. Whaley (1993)described
gression test, he found (consistent with Beckers) the constructionof this index and discussed hedg-
that the nearest-the-money measure contains the ing strategiesbased on futures and options written
most information about future volatility, followed on a volatility index.
by the minimum-squared-pricing-error measure. The VIX is a weighted average of implied

10 FinancialAnalysts Joumal / July-August 1995

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
volatilitiesof eight OEXoption contracts,four calls Rendleman, Schmalensee and Trippi, and Trippi
and four puts, using the two strike prices nearest failed to account for dividends. Chiras and Man-
the money and the nearest two expiration dates. aster, using a sample of 23 monthly observations
Implied volatility is calculated for each option from June 1973 to April 1975, accounted for divi-
using a Cox-Ross-Rubinstein binomial framework dends by converting realized dividends to a con-
that accounts for early exercise and discrete divi- tinuous rate. They found that during the first nine
dends. The weighted average of the eight is con- months of their sample, implied volatility was not
structed by (1) averaging the implied volatilities of significantlybetter than historical standard devia-
the put and the call, holding strike and time to tion at forecastingvolatility. That result, however,
expiration constant; (2) averaging across strike is reversed in the remainderof the sample, leading
prices, weighting in proportion to the distance the authors to conclude that the markettook some
from the strike price to the current index value; time after the opening of the CBOEin 1973before
and (3) averaging across times to expiration with beginning to incorporate volatility forecasts into
weights proportional to the difference between option prices. Melnick and Yannacopoulos com-
time to expiration of the option and 22 days. The pared a volume-weighted implied standard devia-
second step is to standardize the index to repre- tion to a numberof alternativemeasuresof volatility
sent an at-the-money option, and the third is to for IBMcall options over the period November 1,
standardize the index to represent an option with 1976, to July 3, 1980, and concluded that implied
22 trading days to expiration. Fleming, Ostdiek, volatility incorporatesall the relevant information
and Whaley (1995)examined the time-series prop- in past prices. Using data from 1973 to 1981,
erties of this index. Heaton (1986)came to a similar conclusion.
In summary, various weighted-average tech- In short, the early literature found implied
niques for calculatingimplied volatility have been volatilityto be better at forecastingfuture volatility
suggested and have received quite a bit of atten- than estimators based on historical data. Subse-
tion despite widespread admission that the under- quent research has generally supported this con-
lying model used to calculate the volatilities is clusion, but results have been mixed. For one,
incorrectand despite empirical evidence suggest- these early papers generally used the historical
ing that the near-the-moneyoption is as good as a standard deviation of returns based on a time
weighted average at predicting volatility. series of closing prices. Since then, other, more
powerful methods have been developed for pre-
IMPLIED VERSUS
VOLATlLITY dicting future volatility from historical data. One
DATA
HISTORICAL such approachtakes advantage of the information
Perhaps the most important issue in volatility in daily high, low, opening, and closing prices.12
forecastingis whether the forecastshould be based Marsh and Rosenfeld (1986) argued that these
on historicalprice data, implied volatility, or some extreme-value estimators are quite sensitive to
combination of the two. Latane and Rendleman microstructuralfrictionssuch as infrequenttrading
(1976), Schmalensee and Trippi (1978), Chirasand and bid-ask bounce. Beckers (1983) suggested in-
Manaster(1978),Beckers(1981),and several others corporating implied volatility to increase the
all found that implied volatility is better than power of these tests and found that doing so leads
historical standard deviation at forecasting future to relatively small incrementalforecasting power.
realized volatility. Latane and Rendleman found Another approach has been to describe the
this result using 39 weekly observations for op- time series of the underlying stock using general-
tions on 24 stocks from October 1973 to June 1974. ized autoregressive conditional heteroscedasticity
Schmalensee and Trippi used 56 weekly observa- (GARCH)models.13For example, Day and Lewis
tions for options on six stocks from April 1974 to (1992) studied the relative forecasting power of
May 1975. Besides finding no significant relation- implied volatility versus historical data by adding
ship between historicalvolatility and implied vol- implied volatility as an explanatory variable in a
atility, they also found that implied volatility GARCHmodel. They found that for OEXoptions,
seems to decline following price increases and that both implied volatility and historical data contain
implied volatilities are positively correlatedacross incrementalinformationabout future volatility. Xu
stocks. Trippi (1977) described a trading strategy and Taylor (1993) extended this approach to ac-
based on implied volatilities that appears to have count for the term structureof volatilityand found
been capable of earning abnormalreturns. that for three out of four foreign exchange options,
Unfortunately, the papers of Latane and implied volatility is the best one-period predictor

Financial Analysts Journal / July-August 1995 11

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
of volatility and that historical data add no addi- shocks to the implied volatility of near-expiration
tional explanatory power. Choi and Wohar (1993) options should be accompanied by shocks in the
found that returns forecasted from a GARCH same direction (but smaller in magnitude) to the
model are consistent with the implied volatilities implied volatility of long-term options. Stein hy-
observed in option prices. Fung and Hsieh (1991) pothesized an AR(1) process for the volatility of
addressed the forecasting issue in the context of a the OEXand estimated the mean-reversionparam-
GARCHmodel, using transactions-leveldata from eter using historical data. He found that the im-
options on S&P futures, U.S. Treasury bond fu- plied volatility of the long-term OEX options
tures, and deutschemark futures. Of the three, seems to overreactto changes in the implied vol-
implied volatilityhelped forecastthe volatilityonly atility of short-term OEX options, given the ob-
of deutschemark futures. Noh, Engle, and Kane served level of mean reversion in volatility. This
(1994) compared the forecasting ability of implied result conflictswith the earlierwork of Poterbaand
volatility with that of a GARCHmodel by compar- Summers (1986), who examined properties of his-
ing the returns to delta-neutral straddles of S&P torical volatility and of the volatilities implied by
500 options based on the two forecasts. They the CBOEcall option index and by the Value Line
found that a strategy based on the GARCHfore- three-month and six-month options index. They
cast generated higher returns than a similar strat- concluded that volatilityshocks do not persist very
egy based on implied volatility. Strong and Dick- long and that the implied volatility of long-run
inson (1994) also discussed how to use both options does not move very much in response to
historicaland implied volatility to forecastimplied shocks in that of short-run options. The results of
volatility and hedge ratios. Diz and Finucane (1993)supported the findings of
Lamoureux and Lastrapes (1993) used the Poterba and Summers and contradicted those of
GARCHformulationand implied volatilitiesto test Stein. Using an alternativeempiricalspecification,
the Hull-White class of stochastic-volatilityoption Diz and Finucane (1993) found strong evidence
pricing models. These models assume that volatil- against the over-reaction hypothesis and limited
ity risk is unpriced (see Hull and White 1987). If a evidence in favor of under-reactionin some peri-
model in this class is correct, then all available ods.
informationwill be incorporatedinto the market's Using a statistical technique that explicitly
prediction of future volatility.14 The data for the accounts for overlapping observations, Fleming
test include transactions data for CBOE options (1994) studied the forecasting power of implied
tradingon ten actively traded stocks from April 19, volatilityfor OEXoptions.16He calculatedimplied
1982, to March 31, 1984. None of the stocks paid volatility very carefully using a binomial tree that
dividends during this period, and the option incorporated dividends, early exercise, and the
prices used to calculate implied volatilities were embedded "wildcard option."17 Fleming con-
constructed by taking the midpoint of the inside cluded that implied volatility is an upwardly bi-
bid-ask quote for at-the-money, intermediate-term ased estimator of future volatility but that the
options. Out of these data, a single, representative magnitude of the bias is not economically signifi-
daily implied volatility observation was con- cant. He also concluded that implied volatility
structed for each stock, and the resulting time dominates historical volatility as a forecast of fu-
series was subjected to analysis. For seven of the ture volatility.
ten stocks, the authors rejected the Hull-White Canina and Figlewski (1993)also investigated
class of models in favor of a more general GARCH the ability of implied volatility to forecast actual
model.15 Diz and Finucane (1994) examined the volatility, but they came to the opposite conclu-
relationship between implied volatility and the sion. They claimed that implied volatility has vir-
average volatility expected over the remaining life tually no explanatorypower but that estimates of
of options. Using data from OEX options, they historicalvolatility can explain some of the varia-
concluded that, consistent with the Hull-White tion in realized volatility. The authors iterated a
class of option pricing models, implied volatilityis 500-step binomial tree to obtain daily observations
not significantly different from expected average of implied volatility for OEX call options from
volatility for a variety of volatility process specifi- March1983to March1987, excluding options with
cations. fewer than 7 or more than 127 days to maturity.
Stein (1989)used implied volatility to test the They used realized dividend values to adjust for
"overreaction"hypothesis in the options market. dividends, allowed for early exercise, and threw
If volatility follows a mean-revertingprocess, then away option prices that violated the static-arbi-

12 FinancialAnalysts Joumal / July-August 1995

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
trage boundary. They calculated implied volatili- leads the stock market and is negatively related
ties for four "time-to-maturity"groups and eight when the stock market leads the options market.
"intrinsicvalue" groups so as to reduce the effects Lo and Wang (1995)pointed out that because
of systematic volatility-smile effects. They then prices are sampled discretely, an estimate of his-
estimated the realized volatilityover the remaining toricalvolatility will depend on the analyst's belief
life of the option and (for each group) regressed about the level of mean reversion in stock prices.19
that estimate on implied volatility. They found the This result has subtle implications for the debate
regression coefficients to be statistically insignifi- on whether to use implied volatility or historical
cant and so concluded that implied volatility data. On the one hand, it means that measuringan
poorly forecasts actual volatility. A regression of asset's instantaneous volatility using historical
realized volatility on historicalvolatility, however, data is difficult;on the other hand, it also means
appeared to have some explanatorypower. Addi- that using implied volatility to estimate the vari-
tional research on this topic by Geske and Kim ance of the asset's price at a future date is difficult.
(1994) and by Christensen and Prabhala (1994) Perhaps the proper choice of whether to use im-
casts serious doubt on these results. plied volatility or historical data depends on the
Although the debate is still open, the general forecastinghorizon.
conclusion to be drawn from this large body of
research is that for forecasting volatility, implied IMPLIED EVENTSTUDIES
VOLATILITY
volatility tends to be more useful than historical Because implied volatility is widely interpreted as
data. Time-series models that incorporate both, the market's forecast of future volatility, move-
however, show a great deal of promise. A number ments in implied volatility have been interpreted
of recent papers shed more light on the intertem- as reflecting the market's response to new infor-
poral relationship between implied volatility and mation about the future volatilityof the underlying
underlying price dynamics. Sheikh (1993), for ex- stock. This interpretationhas led several authors
ample, examined the time series of implied vola- to conduct event studies examining the impact of
tility and its relationship to returns in the under- new information on the implied volatility of op-
tions.
lying stock for a number of equity and index
Patell and Wolfson (1981)examined the prop-
options trading on the CBOE. He found positive
erties of the implied volatility of equity options at
autocorrelationin the time series of implied vola-
the time of quarterly earnings announcements.
tilities and a positive relationshipbetween returns
Other authors have found historicalvolatilityto be
and lagged implied volatility.
high near earnings announcements. If this volatil-
Kawaller, Koch, and Peterson (1994)used in- ity is reflected in option prices, then implied vola-
traday data to examine the lead-lag relationship tility should fall after earnings announcements.
between implied and historical volatility for op- Patell and Wolfson verified this prediction.
tions on futures. Included in their study were A number of authors have examined the be-
options on S&P 500 futures, deutschemark fu- havior of implied volatility in response to stock
tures, Eurodollarfutures, and live cattle futures. splits. Historicalvolatility tends to be high follow-
Dividing the day into ten intervals, they investi- ing stock splits. Frenchand Dubofsky (1986)found
gated the intraday relationship between implied a small increase in implied volatility in response to
and historical volatility. They found that implied stock splits. This finding is contradicted by the
volatility never leads historical volatility, suggest- results of Klein and Peterson (1988) and Sheikh
ing that option traders cannot anticipate impend- (1989), who found that implied volatility does not
ing changes in volatility. Using daily data, this seem to respond to stock splits.
result reverses in some markets. Their study also Day and Lewis (1988)found that implied vol-
found a strong link between trading volume and atility is higher around the expiration dates of
historical volatility but no stable relationship be- stock index futures and stock index options. Bailey
tween volume and implied volatility.18Boyle and (1988)examined the response of implied volatility
Park (1994) showed that a lead-lag relationship to the release of (Ml) money supply information.
between stock and option markets can lead to a Gemmill (1992) examined the pattern of implied
bias when implied volatility is calculated using volatility in British markets immediately prior to
contemporaneous observations. In particular,the the election of 1987. Madura and Tucker (1992)
size of the bias is positively related to the volatility considered the effect of U.S. balance-of-tradedef-
of the underlying asset when the option market icit announcements on the implied volatility of

Financial Analysts Journal / July-August 1995 13

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
currency options. Levy and Yoder (1993) investi- thinking of trading in terms of Black-Scholes im-
gated the behavior of implied volatility around plied volatility.
merger and acquisition announcements, and Bar- The earliest papers that found evidence for
one-Adesi, Brown, and Harlow (1994) used the volatility smiles did not formulate the results in
implied volatility of options on target firms to such terms but instead described how Black-
estimate the probability of a successful takeover. Scholes pricing errors vary systematically with
Jayaramanand Shastri (1993) examined the rela- strike price or with time to expiration. MacBeth
tionship between implied volatilityand announce- and Merville(1979),for example, reportedthat the
ments of dividend increases. Black-Scholes model undervalues in-the-money
and overvalues out-of-the-money call options.
IMPLIED SMILES
VOLAi1LITY Subsequent authors found the contrary result-
Over the years, it has become quite clear that the that the Black-Scholesmodel undervalues out-of-
market does not price all options according to the the-money calls. These and other relevant results
Black-Scholes formula. The consensus opinion is are summarized by Galai (1983).
that the model performsreasonablywell for at-the- Of the studies documenting volatility smiles,
money options with one or two months to expira- the most systematicand complete is that of Rubin-
tion, and this experience has motivated the choice stein (1985).Rubinsteinexamined matched pairs of
of such options for calculating implied volatility. call option transactionsfrom the BerkeleyOptions
For other options, however, the discrepanciesbe- Data Base to conduct nonparametrictests of the
tween market and Black-Scholes prices are large Black-Scholesnull hypothesis that implied volatil-
and systematic. If the marketwere to price options ities exhibit no systematic differencesacross strike
according to the Black-Scholes model, all options prices or across time to maturity for otherwise
would have the same implied volatility. Because identical options.21 If deviations from the Black-
the Black-Scholesmodel holds reasonablywell for Scholes model are white noise, the option with the
some options and not for others, differentoptions lower strike price should have a higher implied
on the same underlying security must have differ- volatilityfor about half the observations. A similar
ent Black-Scholes implied volatilities. It is now argument applies for the option with the shorter
well known that the implied volatilities of options time to maturity.
differsystematicallyacross strike prices and across Rubinstein's most robust result is that for
time to expiration. The pattern of implied volatili- out-of-the-moneycalls implied volatilityis system-
ties across times to expiration is known as the atically higher for options with shorter times to
"term structure of implied volatilities," and the expiration. His other results were statisticallysig-
pattern across strike prices is known as the "vola- nificant but changed across subperiods. He di-
tility skew" or the "volatilitysmile," a term that is vided the sample into two subperiods: Period I
sometimes used generally to refer to the pattern from August 23, 1976, to October 21, 1977, and
across both time to expiration and strike. Period II from October 24, 1977, to August 31,
To even talk about volatility smiles is schizo- 1978. For at-the-money calls, Rubinstein found
phrenic: First, a constant volatility is assumed to that in Period I, implied volatility for options with
derive the model; then, many differentvolatilities short times to expirationwas higher than for those
are calculatedfor the same underlying asset. Once with longer times to expirationbut that the result
the Black-Scholesmodel is rejected, Black-Scholes was the opposite in Period II. Moreover, in Period
implied volatility has no real meaning and, of I, implied volatility was higher for options with
course, should no longer be interpreted as the lower striking prices, but again, the result was
market'sassessment of the underlying asset's vol- reversed in Period II. Thus, systematic deviations
atility. The real phenomenon underlying volatility from the Black-Scholesmodel appear to exist, but
smiles is that either (1) market imperfections sys- the pattern of deviations varies over time.22
tematically prevent prices from taking their true Subsequent studies by Sheikh (1991) and by
Black-Scholes values or (2) the underlying asset Heynen (1994) used Rubinstein's nonparametric
price process differsfrom the (lognormaldiffusion) tests to examine implied volatilitypatternsin index
process assumed by the Black-Scholes model.20 options. Sheikh found smile effects using transac-
The volatility smile is just a convenient way of tions data for OEXcall options from March1983to
illustrating this observation that probably devel- March1987. He observed that the smiles constitute
oped by historicalaccident-motivated by the fact evidence against the Black-Scholes model and in
that options traders have grown accustomed to favor of an option pricing model that incorporates

14 FinancialAnalysts Journal / July-August 1995

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
stochasticvolatility. Heynen examined the implied either allow for volatility smiles or explicitly incor-
volatility of European Options Exchange (EOE) porate them into the option pricing process.
stock-index options, which are European style One approach is to use stochastic volatility
options on an index of 25 active stocks on the models, the focus of one of the most productive
Amsterdam Stock Exchange. Using Rubinstein's research areas in option pricing during the past
approachand transactionsdata from January23 to few years. If the price of the underlying asset is
October 31, 1989, he found systematic smile ef- assumed to follow a sufficiently complicated sto-
fects, including a U-shaped term structure of im- chastic process, nearly any type of volatility smile
plied volatility. He reviewed the predictions of can be generated.
various stochastic volatility models, found the ob- Another approach is to use the information
served smile pattern to be inconsistent with them, contained in option prices to estimate the risk-
and suggested an alternative explanation, based neutral density of the terminal stock price. This
on market imperfections, for the volatility smile. method relies on a general result from modern
Many other authors found evidence for vola- option pricing theory: Under certain conditions, a
tility smiles and nonflat term structuresof implied contingent claim that depends on the terminal
volatilitiesin various markets. Shastriand Tandon stock price (and that cannot be exercised early)can
(1986), for example, used the Geske-Johnson ap- be priced by describingthe contractas a bundle of
proach (see Geske and Johnson 1984) to price state-contingent claims, multiplying the payoff in
American options on futures and found volatility- each state by the corresponding "Arrow-Debreu"
smile and term-structureeffects in the marketsfor state price, and summing across states.23 Thus,
options on S&P 500 futures and on deutschemark given N different states, the time t price of a
futures. They also suggested using yesterday's contingent claim expiring at time T would be
estimate of implied volatility as an input for to- calculatedby the equation
day's option pricing model and examined the
performance of this approach relative to using N

historicalvolatility. Xu and Taylor(1994)examined C(t)= E V(s)p(s), (6)

s=1
the term structureof volatility implied by options
on four PhiladelphiaStock Exchangecurrencyop- where V(s) describes the payout at time T and p(s)
tions using data from 1985 to 1989. Heynen, the Arrow-Debreu price of state s. For simplicity,
Kemna, and Vorst (1994) examined the ability of suppose that the risk-free rate, denoted by r, is
various GARCH models to explain the observed constant. Because the owner of a complete set of
term structureof implied volatilities. state-contingentclaims is guaranteed one dollar at
In short, the discrepancy between market expiration,the sum of the state prices must then be
prices and the predictions of the Black-Scholes e-r(T-t), the value of one dollar discounted to the
model has attracted a great deal of interest. But risk-freerate. If equation (6) is rewritten as
again, looking at Black-Scholesvolatility smiles is
awkward-in essence, we must assume that they N p(s)
do not exist in order to derive them. The next C(t)= er(T-t)V(s)(Tt) (7)
section of this article describes another, more ad- s=1
equate method for dealing with the observed fail-
N
ures of the Black-Scholes model.
e- r(T-t)V(S)T(S),
s=1
THEPROBABILITY IMPLIED
DISTRIBUTION
BYOPllON PRICES then the ir(s)sum to 1. Because neither state prices
Although volatility smiles have only recently come nor discount rates can be negative, each wT(s) is
to the attention of the academic community, their non-negative, and the set of ir(s)terms has the two
existence and even their shape have long been essential properties of a probabilitydensity.
familiarto savvy individuals and institutions famil- In fact, if market participantswere risk neu-
iar with options markets. The reality of volatility tral, then for each state, these i(s) terms would be
smiles has led to general dissatisfaction with the the same as the objective probabilityof that state.
Black-Scholes and other models that are inconsis- In other words, the set of state-contingent prices
tent with them. Recent developments in the aca- dividedby the discountfactorwouldsimplybe the
demic literature(and certainlywithin the realm of underlying probability density. This relation is
proprietaryresearch)have considered models that why the iT(s)terms are often called "risk-neutral

Financial Analysts Journal / July-August 1995 15

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
probabilities,"which together form the "risk-neu- the warrants. A good deal of theoretical and em-
tral density." piricalresearch remains to be done on this topic.
When the state space is continuous, the price Rubinstein(1994)discussed three methods for
of a contingent claim is derived by integratingthe estimating the risk-neutraldensity implied by op-
payoff over a density function (the risk-neutral tion prices. The first, attributed to Francis Long-
density) of the underlying asset and then discount- staff, is simply to derive a step-function approxi-
ing at the risk-freerate. The price of a contingent mation to the risk-neutraldensity, where the step
claim at time t is given by the equation function is as coarse as the interval between suc-
cessive strike prices of traded options (five dollars
C(t)- e-r(T-t)J V(s)f(s)ds, (8) for most options traded in the United States).
Rubinstein showed that for some parameter val-
where f(s) is the risk-neutral density. For a call ues, this method yields poor results.25
option, V(s) = max(O,s - K), and for a put option, A second method, introduced by Shimko
V(s) = max(O,K - s), where K represents the strike (1991), relies on the fact that the risk-neutralden-
price. sity is equal to the second derivative of the call
The traditionalway to implement this result is price with respect to the strike price.26Thus, if a
to impose a restriction such as the Black-Scholes continuum of option prices were observable, the
(constant-variance)assumption on the underlying risk-neutraldensity would also be observable. In
asset process and to use the resulting density f(s) to fact, option prices are only observable for a few,
price options. The new approach, suggested in discretely spaced strike prices. Shimko suggested
recent papers by Rubinstein (1994) and other au- estimating the option price as a continuous func-
thors cited therein, is to start with the market tion of the strikepriceby interpolatingthe prices of
prices of options, then find some density f(s) that is market-tradedoptions, then deriving the risk-neu-
consistent with those prices, given the pricing tral density from the interpolated values. Of
equation (8). The variance of this market-implied course, call option prices can be interpolated in
distribution may then be used as a measure of many ways. Shimko's approach is to calculatethe
future volatility over the remaining life of the Black-Scholes implied volatility for each option,
option.24Potentially, the implied risk-neutraldis- then use least squares to fit a quadraticfunction to
tribution contains even richer information about
the volatility smile. The fit values from the least
the market'sexpectations for future movements in
the underlying asset. This distributioncan then be squares procedure give implied volatility as a con-
used, for example, to calibratea binomial or trino- tinuous function of the strike price, which may
mial tree that is consistent with the observed prices then be mapped back through the Black-Scholes
of all options. Methods of incorporatingthe vola- equation to obtain a continuum of call option
tility smile into tree-based option pricing models prices, which in turn yields the risk-neutralden-
have been suggested by Rubinstein (1994), Der- sity.
man and Kani (1994), and Dupire (1994). Rubinstein (1994) suggested a third method,
Kuwahara and Marsh (1994) used Rubin- which is to choose the distribution that is closest
stein's method in their investigation of Japanese (in a least-squaressense) to some "prior"distribu-
equity warrantpricing. They found that the prob- tion, subject to the constraint that the rational
ability density implied by Nikkei index options is option prices derived from the chosen distribution
negatively skewed, as is the case for S&P index must fall within the observed bid-ask spread for
options, but that the density implied by a cross- each traded option. A comprehensive investiga-
section of warrants is positively skewed. Equiva- tion of the merits of alternativepriors or objective
lently, the volatility smile seems to be downward functions constitutes a worthy agenda for further
sloping for index options and upward sloping for research.27

16 FinancialAnalystsJoumal/ July-August1995

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
BIBLIOGRAPHY
Amin, Kaushik I., and Andrew J. Morton. 1994. "Implied Cox, John C., Stephen A. Ross, and Mark Rubinstein. 1979.
VolatilityFunctions in Arbitrage-FreeTerm StructureMod- "OptionPricing:A SimplifiedApproach."Journalof Financial
els." Journalof Financial Economics, vol. 35, no. 2 (April):141- Economics,vol. 7, no. 3 (September):229-63.
80. Culumovic,Louis, and RobertL. Welch. 1994. "A Reexamina-
Bailey, Warren.1988. "Money Supply Announcementand the lion of Constant-VarianceAmerican Call Mispricing."Ad-
Ex Ante Volatility of Asset Prices." Journalof Money, Creditand vances in Futures and Options Research,vol. 7, no. 1:177-221.
Banking,vol. 20, no. 4 (November):611-20. Day, TheodoreE., and CraigM. Lewis. 1988. "TheBehaviorof
Ball, CliffordA., Walter N. Torous, and Adrian E. Tschoegl. the VolatilityImplicitin the Prices of Stock Index Options."
1985. "On InferringStandardDeviations from Path Depen- Journalof Financial Economics, vol. 22, no. 1 (October/Decem-
dent Options." Economics Letters,vol. 18, no. 4:377-80. ber):103-22
Barone-Adesi,Giovanni, Keith C. Brown, and W.V. Harlow. . 1992. "Stock Market Volatility and the Information
1994. "On the Use of ImpliedStockVolatilitiesin the Predic- Content of Stock Index Options." Journalof Econometrics,vol.
tion of Successful CorporateTakeovers."Advancesin Futures 52, nos. 1/2 (April/May):267-87.
and Options Research,vol. 7, no. 1:147-65. Derman,Emanuel,and IrajKani. 1994. "Ridingon the Smile."
Beckers, Stan. 1981. "StandardDeviations Implied in Option RISK,vol. 7, no. 2 (February):32-39.
Pricesas Predictorsof FutureStockPriceVariability."Journal Diz, Femando, and Thomas Finucane. 1993. "Do the Options
of BankingandFinance,vol. 5, no. 3 (September):363-82. Markets Really Overreact?" TheJournalof FuturesMarkets,vol.
. 1983. "Variancesof Security Price Returns Based on 13, no. 3 (May):299-312.
High, Low, and Closing Prices." TheJournalof Business,vol. . 1994. "RationalExpectationsand the RelationshipBe-
56, no. 1 (January):97-112. tween Spot and ImpliedVolatility."Workingpaper, Syracuse
Black, Fischer, and Myron Scholes. 1973. "The Pricing of
University.
Options and CorporateLiabilities."Journalof PoliticalEcon-
Dupire, Bruno. 1994. "Pricingwith a Smile." RISK,vol. 7, no.
omy,vol. 81, no. 3 (MaylJune):637-54.
1 (January):18-20.
Bodurtha,James N., Jr., and Qi Shen. 1994. "ImpliedCovari-
Eckardt,WalterL., Jr., and Stephen L. Williams. 1984. "The
ance in PHLXDeutschemarkand Yen OptionValues."Work-
Complete Options Indexes." Financial Analysts Journal, vol.
ing paper, University of Michigan.
Bollerslev,Tim. 1986."GeneralizedAutoregressiveConditional 40, no. 4 Uuly/August):48-57.
Heteroskedasticity." lournal of Econometrics, vol. 31, no. 3 Edey, Malcolm,and GrahamElliot. 1992. "Some Evidence on
(April):307-27. Option Prices as Predictorsof Volatility."OxfordBulletinof
Boyle, Phelim P., and Hun Y. Park.1994. "ImpliedVolatilityin Economics and Statistics,vol. 54, no. 4:567-78.
Option Pricesand the Lead-LagRelationBetween Stockand Engle, Robert F., and Chowdhury Mustafa. 1992. "Implied
Option Prices." Working paper OFOR94-01, University of ARCH Models from Options Prices." Journal of Econometrics,
Illinois, Urbana-Champaign. vol. 52, nos. 1/2 (April/May):289-311.
Breeden,Douglas T., and RobertH. Litzenberger.1978."Prices Fleming,Jeff. 1994. "TheQuality of MarketVolatilityForecasts
of State-ContingentClaims Implicitin Options Prices." 7he Implied by S&P 100 Index Option Prices." Workingpaper,
Journalof Business, vol. 51, no. 4 (October):621-51. Jones GraduateSchool, Rice University.
Brenner,Menachem, and Dan Galai. 1981. "The Propertiesof Fleming, Jeff, BarbaraOstdiek, and Robert E. Whaley. 1995.
the Estimated Risk of Common Stocks Implied by Option "PredictingStock MarketVolatility:A New Measure." The
Prices."Workingpaper #112, Universityof California-Berke- Journalof Futures Markets, vol. 15, no. 3 (May):265-302.
ley. Franks,Julian R., and Eduardo S. Schwartz. 1991. "The Sto-
. 1982. "On the Prediction of the Implied Standard chasticBehaviourof MarketVarianceImpliedin the Pricesof
Deviation." Workingpaper, Hebrew University,Jerusalem. Index Options." The EconomicJournal,vol. 101, no. 409
Brenner, Menachem, and Marti G. Subrahmanyan.1988. "A (November):1460-75.
Simple Formula to Compute the Implied Standard Devia- French, Dan W., and David A. Dubofsky. 1986. "StockSplits
tion." Financial Analysts Journal, vol. 44, no. 5 (September/ and Implied Stock Price Volatility." TheJournalof Portfolio
October):80-83. Management, vol. 12, no. 4 (Summer):55-59.
Canina, Linda, and Stephen Figlewski. 1993. "The Informa- Fung, Hung-Gay, Chin-JenLie, and Abel Moreno. 1990. "The
tional Content of Implied Volatility."TheReviewof Financial ForecastingPerformanceof the Implied StandardDeviation
Studies, vol. 6, no. 3:659-81. in Currency Options." Managerial Finance, vol. 16, no. 3:24-
Chiras, Donald P., and Steven Manaster.1978. "The Informa- 29.
tion Content of Option Prices and a Test of MarketEfficien- Fung, William K.H., and David A. Hsieh. 1991. "Empirical
cy." Journal of Financial Economics, vol. 6, nos. 2/3 (June/ Analysis of Implied Volatility:Stocks, Bonds, and Curren-
September):213-34. cies." Workingpaper, Duke University.
Choi, Seungmook, and Mark E. Wohar. 1993. "ImpliedVola- Galai, Dan. 1983. "A Survey of EmpiricalTests of Option-
tility in Options Marketsand ConditionalHeteroscedasticity Pricing Models." In Option Pricing: Theory and Applications,
in Stock Markets." The Financial Review, vol. 27, no. 4 (No- MenachemBrenner,(ed.), 45-80. Lexington,Mass.: Lexing-
vember):503-30. tonBooks.
Christensen, Bent Jesper, and N.R. Prabhala.1994. "On the Garman,MarkB., and MichaelJ. Klass. 1980. "On the Estima-
Dynamics and InformationContent of Implied Volatility:A tion of Security Price Volatilitiesfrom HistoricalData." The
Bivariate Time Series Perspective." Working paper, New Journalof Business,vol. 53, no. 1 (January):67-78.
YorkUniversity. Garman,Mark B., and Steven W. Kohlhagen. 1983. "Foreign

Financial Analysts Joumal / July-August 1995 17

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
 Currency Option Values." Journal of InternationalMoney and the Black-ScholesModel FitJapaneseWarrantsand Convert-
Finance,vol. 2, no. 3 (December):231-37. ible Bonds?"Japanese
Journalof FinancialEconomics,
vol. 1, no.
Gastineau, Gary L. 1977. "An Index of Listed Option Premi- 1.
ums." FinancialAnalysts Journal, vol. 33, no. 3 (May/June):70- Lamoureux,ChristopherG., and WilliamD. Lastrapes. 1993.
75. "ForecastingStock-ReturnVariance:Towardan Understand-
Gastineau, Gary L., and Albert Madansky. 1979. "Why Simu- ing of StochasticImplied Volatilities."TheReviewof Financial
lations Are an UnreliableTest of Option Strategies."Financial Studies,vol. 6, no. 2:293-326.
AnalystsJournal,vol. 35, no. 5 (September/October):61-76. Latane, Henry A., and RichardJ. Rendleman,Jr. 1976. "Stan-
. 1984. "Some Comments on the CBOE Call Options dard Deviations of Stock Price Ratios Implied in Option
Index." FinancialAnalysts Journal,vol. 40, no. 4 Uuly/August): Prices." TheJournalof Finance,vol. 31, no. 2 (May):369-81.
58-67. Levy, Haim, and James A. Yoder. 1993. "The Behavior of
Gemmill,Gordon. 1986."TheForecastingPerformanceof Stock Option Implied Standard Deviations Around Merger and
Options on the London Traded Options Market."Journalof AcquisitionAnnouncements." TheFinancialReview,vol. 28,
Business Finance and Accounting, vol. 13, no. 4 (Winter):535- no. 2 (May):261-72.
46. Lo, Andrew W., and JiangWang. 1995. "ImplementingOption
.1992. "PoliticalRiskand MarketEfficiency:Tests Based Pricing Models When Asset Returns are Predictable."The
on BritishStock and Option Marketsin the 1987 Election." Journalof Finance,vol. 50, no. 1 (March):87-129.
Journalof Banking and Finance, vol. 16, no. 1 (March):43-73. MacBeth,J., and L. Merville.1979. "An EmpiricalExamination
Geske, Robert, and H.E. Johnson. 1984. "The American Put of the Black-ScholesCall Option PricingModel." TheJournal
Option Valued Analytically."TheJournalof Finance,vol. 39, of Finance,vol. 34, no. 5 (December):1173-86.
no. 5 (December):1511-24. Madura, Jeff, and Alan L. Tucker. 1992. "TradeDeficit Sur-
Geske, Robert, and Kwanho Kim. 1994. "RegressionTests of pluses and the Ex-Ante Volatility of Foreign Exchange
Volatility Forecasts Using EurodollarFutures and Option Rates."Journalof International MoneyandFinance,vol. 11, no.
Contracts." ASGM working paper #31-93, University of 5 (October):492-501.
California-LosAngeles. Maloney, Kevin, and Richard Rogalski. 1989. "Call-Option
Harvey, CampbellR., and RobertE. Whaley. 1991."S&PIndex Pricingand the Tum of the Year."TheJournalof Business,vol.
Option Volatility." The Journal of Finance, vol. 46, no. 4 62, no. 4 (September):539-52.
(September):1551-62. Manaster,Steven, and GaryKoehler.1982."TheCalculationof
Heath, David, Robert Jarrow, and Andrew Morton. 1992. Implied Variancesfrom the Black-ScholesModel: A Note."
"Bond Pricing and the Term Structureof Interest Rates: A TheJournalof Finance,vol. 37, no. 1 (March):227-30.
New Methodologyfor ContingentClaimsValuation."Econo- Marsh, TerryA., and EricR. Rosenfeld. 1986. "Non-Trading,
metrica,vol. 60, no. 1 (january):77-105. Market Making, and Estimates of Stock Price Volatility,"
Heaton, Hal. 1986. "VolatilitiesImplied by Options Premia:A Journalof FinancialEconomics, vol. 15, no. 3 (March):359-72.
Test of MarketEfficiency."TheFinancialReview,vol. 21, no. 1 Mayhew, Stewart.1995."On Estimatingthe Risk-NeutralProb-
(February):37-49. ability Distribution Implied by Option Prices." Working
Heynen, Ronald. 1994. "An EmpiricalInvestigation of Ob- paper, Universityof California-Berkeley.
served Smile Patterns."Workingpaper, TinbergenInstitute, Melnick, Edward L., and Dimitri Yannacopoulos. Undated.
ErasmusUniversity, Rotterdam. "An EmpiricalInvestigationof Estimatorsof Stock Returns
Heynen, Ronald, Angelica Kemna, and Tom Vorst. 1994. Volatility."Workingpaper, New YorkUniversity.
"Analysis of the Term Structure of Implied Volatilities." Morse, Joel N. 1991. "An IntraweekSeasonalityin the Implied
Journal of Financial and Quantitative Analysis, vol. 29, no. 1 Volatilitiesof Individual and Index Options." TheFinancial
(March):31-56. Review,vol. 26, no. 3 (August):319-41.
Hull, John, and Alan White. 1987. "The Pricingof Options on Noh, Jaesun, RobertF. Engle, and Alex Kane. 1994. "Forecast-
Assets with StochasticVolatilities."TheJournalofFinance,vol. ing Volatility and Option Prices of the S&P Index." The
42, no. 2 (une):281-300. Journalof Derivatives,vol. 2, no. 1 (Fall):17-30.
Jayaraman,Naryanan,and Kuldeep Shastri.1993. "TheEffects Parkinson,M. 1980. "The ExtremeValue Method For Estimat-
of the Announcements of Dividend Increaseson Stock Vol- ing the Variance of the Rate of Return." The Journalof
atility: The Evidence from the Options Market."Journalof Business,vol. 53, no. 1 Uanuary):61-65.
Business Finance and Accounting, vol. 20, no. 5:673-85. Patell, James M., and Mark A. Wolfson. 1981. "The Ex Ante
Kawaller,Ira G., Paul D. Koch, and John E. Peterson. 1994. and Ex Post Price Effectsof QuarterlyEarningsAnnounce-
"Assessing the IntradayRelationshipBetween Implied and ments Reflected in Option and Stock Prices." Journalof
HistoricalVolatility."TheJournalof FuturesMarkets,vol. 14, AccountingResearch,vol. 19, no. 2 (Autumn):434-58.
no. 3 (May):323-46. Poterba, James M., and Lawrence H. Summers. 1986. "The
Klein, Linda, and David R. Peterson. 1988. "InvestorExpecta- Persistenceof Volatilityand StockMarketFluctuations."The
tions of Volatility Increases Around Large Stock Splits As AmericanEconomic Review,vol. 76, no. 5 (December):1142-51.
Implied in Call Option Premia." The Journalof Financial Resnick, Bruce G., Amir Sheikh, and Yo-Shin Song. 1993.
Research,vol. 11, no. 1 (Spring):71-86. "TimeVaryingVolatilitiesand Calculationof the Weighted
Kritzman, Mark. 1991. "What PractitionersNeed To Know Implied StandardDeviation."Journalof Financialand Quanti-
About Estimating Volatility: Part I." FinancialAnalysts Journal, tativeAnalysis,vol. 28, no. 3 (September):417-30.
vol. 47, no. 4 (July/August):22-25. Rubinstein, Mark. 1985. "NonparametricTests of Alternative
Kunitomo,Naoto. 1992. "Improvingthe ParkinsonMethod of Option Pricing Models Using All Reported Trades and
EstimatingSecurityPriceVolatilities."TheJournalof Business, Quotes on the 30 Most Active CBOEOption Classes from
vol. 65, no. 2 (April):295-302. August 23, 1976, through August 31, 1978." TheJournalof
Kuwahara,Hiroto, and TerryA. Marsh. 1994. "Why Doesn't Finance,vol. 40, no. 2 (June):455-80.

18 FinancialAnalysts Joumal / July-August 1995

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
 . 1994. "Implied Binomial Trees." TheJournalof Finance, . 1993. "Bounds of Probability."RISK,vol. 6, no. 4:33-
vol. 49, no. 3 (uly):771-818. 37.
Rubinstein, Mark, and Anand M. Vijh. 1987. "The Berkeley Stein, Jeremy. 1989. "Overreactionsin the Options Market."
Options Data Base:A Tool for EmpiricalResearch."Advances TheJournalof Finance,vol. 44, no. 4 (September):1011-23.
in Futures and Options Research,vol. 2, no. 1:209-21. Strong, Robert A., and Amy Dickinson. 1994. "Forecasting
Schmalensee, Richard, and Robert Trippi. 1978. "Common BetterHedge Ratios."FinancialAnalystsJournal,vol. 50, no. 1
StockVolatilityExpectationsImpliedby Option Premia."The (January/February):70-72.
Journalof Finance, vol. 33, no. 1 (March):129-47. Trippi, Robert R. 1977. "A Test of Option Market Efficiency
Scott, Elton, and Alan L. Tucker. 1989. "PredictingCurrency Using a Random-WalkValuationModel."Journalof Economics
Return Volatility." Journalof Bankingand Finance, vol. 13, no.
andBusiness,vol. 29, no. 2 (Winter):93-98.
6 (December):839-51.
Turvey, Calum G. 1990. "AlternativeEstimates of Weighted
Shastri, Kuldeep, and Kishore Tandon. 1986. "An Empirical
Implied Volatilitiesform Soybean and Live Cattle Options."
Test of a ValuationModel for AmericanOptions on Futures
Contracts." Journal of Financial and Quantitative Analysis, vol.
TheJournalof FuturesMarkets,vol. 10, no. 4 (Winter):353-56.
21, no. 4 (December):377-92. Whaley, RobertE. 1982. "Valuationof AmericanCall Options
Sheikh, Aamir. 1989. "Stock Splits, Volatility Increase, and on Dividend-Paying Stocks: Empirical Tests." Journalof
Implied Volatilities." The Journalof Finance,vol. 44, no. 5 FinancialEconomics, vol. 10, no. 1 (March):29-58.
(December):1361-72. . 1993. "Derivativeson MarketVolatility:Hedging Tools
. 1991. "TransactionData Tests of S&P 100 Call Option Long Overdue." ChicagoBoardOptionsExchangeRiskManage-
Pricing." Journalof Financialand QuantitativeAnalysis, vol. 26, mentSeries.
no. 4 (December):727-52. Xu, Xinzhong,and Stephen Taylor.1993. "ConditionalVolatil-
. 1993. "The Behavior of Volatility Expectations and ity and the InformationalEfficiencyof the PHLXCurrency
TheirEffecton ExpectedReturns."TheJournalofBusiness,vol. Options Market." Working paper, Financial Options Re-
66, no. 1 (january):93-116. search Centre, Universityof Warwick.
Shimko, David. 1991. "Beyond Implied Volatility:Probability . 1994. "The Term Structure of Volatility Implied by
Distributionsand Hedge Ratios Implied by Option Prices." ForeignExchangeOptions."Journalof FinancialandQuantita-
Workingpaper, Universityof SouthernCalifornia. tiveAnalysis,vol. 29, no. 1 (March):57-74.

FOOMhOTES
1. It is importantto understand,however, that averageinstan- Whaley (1991)demonstratedthe potential for microstruc-
taneousvolatilityis not the same thing as totalvarianceover turaldata imperfectionssuch as bid-ask spreadsand asyn-
the remaininglife of the option. chronous observationof option prices and the underlying
2. For example, Engle and Mustafa(1992)showed how to use stock priceto bias implied volatilityestimatesand to induce
option prices to estimatethe volatilityparameterswhen the spurious negative autocorrelationin the time series of
underlying asset follows a GARCHprocess. implied volatility.
3. See the paper by Bodurthaand Shen (1994),who used the 10. The Garman-Kohlhagenformula is similar to the Black-
prices of currency options not only to calculate implied Scholes formula with the foreign interest rate acting as a
volatilitiesfor two exchange rates but also to estimate the continuous dividend yield.
implied correlationbetween them. 11. For additionaldiscussion on this issue, see the papers by
4. Amin and Morton (1994), for example, used Eurodollar Gastineau and Madansky (1979, 1984) and Eckardtand
futuresand options data to backout the impliedvolatilityof Williams(1984).
interest rates in the Heath-Jarrow-Mortonterm-structure 12. This branchof literatureincludes the papers of Parkinson
model. (1980),Garmanand Klass (1980),and Kunitomo(1992).
5. In general, a, may be a vector of parametersdescribingthe 13. The GARCHclass of time-seriesmodels allows the variance
underlying stock process. of the innovation to depend on lagged innovations and
6. The bisectionmethod is to bracketthe root, then repeatedly lagged levels of the process and to revert to a long-run
cut the bracketin half to converge on the root. mean. See Bollerslev(1986).
7. The Newton-Raphson root-findingmethod speeds up con- 14. Technically,the market'svariance forecast error at time t
vergence by taking advantage of informationin the func- should be orthogonalto any informationavailableat time t.
tion's first derivative. Manaster and Koehler (1982) de- 15. One potentialcriticismof this type of test is that biases can
scribedhow to choose a startingvalue for the firstiteration arise from the facts that (1) the implied variancefrom the
to ensure that the Newton-Raphson algorithm will con- Hull-White type models is only an approximationto the
verge whenever a solution actually exists. For a review of true subjective implied variance, and (2) the Brownian
how to apply both bisection and Newton-Raphson meth- motions correspondingto the stock process and the vola-
ods, see Kritzman(1991). tility process may be correlated.To address these potential
8. The option with the highest vega is usuallyvery close to the biases, the authorspresenteda simulationthat investigated
money. the effect of such misspecificationon the results. They
9. Brenner and Galai (1982) also examined the time series found that the bias is never more than 1.3 percent of the
propertiesof these implied volatilityestimates. Harveyand true variance.

Financial Analysts Journal / July-August 1995 19

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions
16. Specifically,he used generalized method of moments esti- claim that pays one dollar if the state occurs and zero
mation. otherwise. In this case, the "states"correspondto different
17. The wildcardoption is the option to exercisean option after terminalvalues of the underlying stock.
the settlement price is fixed. For example, OEX options 24. Remember,however, that this market-implieddistribution
may be exercised until 3:15 (CST),but the settlement price is not the varianceof the true probabilitydistributionbut
is fixed at 3:00. rather of the distribution in a risk-neutral world. The
18. The authors also found a strong U-shaped pattern in theoretical relationship between the two, however, de-
intraday historical volatility but no significant pattern in
pends on the equilibriumprice process in the economy,
intradayimplied volatility.
19. More generally, with discretelysampled data, the estimate which in turn depends on investor preferences.
of the diffusion term depends on the drift term. 25. Mayhew (1995)generalizedthis approachby introducinga
20. Equivalently, the "risk-neutral"density function of the class of spline estimators.
terminalstock price is not lognormal. 26. This fact was first demonstrated by Breeden and Litzen-
21. For a description of the Berkeley Options Data Base, see berger (1978).See also Shimko (1993).
Rubinsteinand Vijh (1987). 27. I would like to thank David Modest for guidance and
22. For a similarstudy using more recent data, see Culumovic support. I would also like to thank MarkRubinstein,Terry
and Welch (1994). Marsh, Bill Keirstead,and Van Harlow for valuable com-
23. The Arrow-Debreu price is the price of a state-contingent ments.

A-k
b

1 __~~~~~~~~~~~~~~~~~~~~~~~~~~--

MONTREAL,CANADA
Forinformation
andregistration: Dubois
Jean-Pierre (514)982-0308
communications
at Enigma

20 FinancialAnalystsJournal/ July-August1995

This content downloaded from 142.51.1.212 on Fri, 25 Oct 2013 01:27:44 AM

All use subject to JSTOR Terms and Conditions