Вы находитесь на странице: 1из 5

CFA Institute

What Does an Option Pricing Model Tell Us about Option Prices? Author(s): Stephen Figlewski Source: Financial Analysts Journal, Vol. 45, No. 5 (Sep. - Oct., 1989), pp. 12-15 Published by: CFA Institute

Accessed: 20/10/2010 17:01

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=cfa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

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.

information about JSTOR, please contact support@jstor.org. CFA Institute is collaborating with JSTOR to digitize,

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


Current Issues:


What Does an Option Pricing Model Tell Us About Option Prices?

by StephenFiglewski,Professorof Fi- nance,Stern Schoolof Business,New YorkUniversity*

The story is told of a Seekerof Knowl- edge who sets off in search of the answer to a question that has troubled him for a long time. In his travels he hearsof two wise men who aresaid by many to be very knowledgeable and experiencedin such matters. The first,a famous guru, lives at the top of a mountain, high above the hustle and bustle of everyday life. Af- ter a strenuous climb, the Seeker is able to pose his question: "Whatis a call option worth?" The guru answers immediately, "It is not hard to prove that

C =


[d] -

X e-rT N

[d -



the stock price, X the strike

where S is

price, T time to expiration,r the risk- free interest rate, crvolatility, N[] de-

notes the cumulativenormal distribu- tion and

d = (log(S/X)+ (r + oJ12)T)/IoHT.

Of course, this has to be modified somewhat in practiceto take into ac- count dividends, the value of early exercise and a few other technicalde- tails (see the appendix)." This answer seems pretty exact, if a bitcomplicated.The Seekerthanksthe guru warmly and goes on his way. The second wise man lives in the middle of a city, surrounded by a continuous swirl of noise and activity. Once the Seeker is able to get his attention,he poses the questionagain:

"Whatis a call option worth?" Again the answer is immediate:

* TheauthorthanksFischerBlack,Stephen Brown,JoelHasbrouck,MarkRubinstein and WilliamSilberfor commentson an earlierdraftof thispaper.

"Thatdepends. Are you buying them

or selling them?" Not knowing quite what to say, the

respond by repeating

Seeker starts to

the words, and equations, of the first

guru, but he is quickly interrupted with: "I don't care about all of that stuff.Tellhim to makeme a bid. Then we can talkaboutwhat a calloption is reallyworth." Somewhat confused and not at all sure the wise men's answers have brought him any closer to enlighten- ment, the Seeker goes away to medi- tate furtheron his question.

Valuation Models and Pricing Models

This parableoffers two very different answers to the same basic question. The distinctionbetween them reflects the not often recognizeddifferencebe- tween theories of option valuationand option pricing. The Black-Scholesmodel and others likeit aretheoriesthattryto derivethe valueof an option so that it is consis- tent with the price of the underlying stock. They assume a marketenviron- ment in which a dynamic riskless ar- bitragestrategywith the stockand the option is possible, and find the value of the option as a component of the arbitrageportfolio. In this ideal market,if the option's price should differ from the model value, an arbitrageur can trade it againstthe correctnumberof sharesof

stock to produce a position that is riskless over the next instant of time.

Continuous rebalancing keeps

hedged position riskless until the op- tion expirationdate. Butits returnwill be higher than the risk-free interest rateby the exactamountby which the option was mispricedat the outset. As there are no constraintson the size of their positions, arbitrageurswill offer an unlimitednumberof options at any


price above the theoreticalvalue and will have infinite demand at any price below it, so optionpricesin the market are driven to their model values. This is the reasoningbehind the firstguru's answer. But in trying to apply a theoretical valuationmodel to the realworld, it is immediately clear that none of the model assumptions actually holds. The arbitragestrategy, which is risk- less and costless in theory, is neither in practice.There is risk because the position can't be rebalancedcontinu- ously when markets are closed, and there are costs because even less- than-continuousrebalancingcan lead to large transaction costs. Even the theoreticaloption value itself is uncer- tain, because it depends on the vola- tility of the stock, which cannot be known exactly. Unlimited arbitrage does not dominate the market. In actualmarkets,option prices,like prices for everything else, are deter- mined by supply and demand. This includes supply and demand from non-arbitrageurs. Investors demand call options because the options offer participationin stockpricemovements on the upside, limitriskon the down- side, and allow investors to control a large amount of stock with a small investment. Call writers supply call options to the marketbecause it is a way to generate income when they expect stockpricewill not rise sharply in the near future. Option demand and supply are also influenced by the market environ- ment, whichencompassestaxes,trans- action costs, margin treatmentof dif- ferent securities, delivery features of option contracts, constraints on mar- gin purchases and short sales of the stock,interactionbetween options and related futures contracts, and many other things. Anything that affectsin- vestors' tradingdecisions but is not in


the model tends to push the market price away from the model value. As the second wise man indicated, a call option is worth exactly the price at


and that does not depend on just the model. A skepticmight ask:Withoutunlim- ited arbitragebetween the option and its underlying stock as a foundation, how can a theoreticalvaluationmodel tell us anythingabout actual market prices? One possible answer is that, under certainconditions, the equilibriumop- tion price is the Black-Scholesvalue even when there is no arbitrage,be- cause when investors evaluate the op- tion as if it were any otherasset, thatis what its payoff patternis worth.1Un- fortunately,one of the necessary con- ditions for this result is that all inves- torsbe identical,which is no moretrue of real markets than the continuous arbitrageassumptions it replaces. An argument that is on stronger ground, but has weaker implications, is that as long as the arbitrageis pos- sible, it will be done in spite of trans- action costs and risk whenever the expectedprofitis largeenough to com- pensate for them. This leads to arbi- tragebounds around the model value. Withinthese bounds, there is no arbi- trage and market price can move

which it can be traded in the

freely, but if the price strays too far from the model value, arbitragebe- comes profitableand will tend to push price back into the bounded range. How much information the model actuallygives us about what the mar- ket pricewill be depends on how wide the arbitragebounds are. This is not

easy to determine, because the trans- action costs and risk for the arbitrage are a function of which random path the stock price follows. In a recent paper, I simulated a large number of price paths and discovered that the arbitrage bounds are disturbingly

wide, even

ample, the price of a one-month, at- the-money call with a Black-Scholes value of $2.05could be anywhere from $1.74 to $2.35 without giving an arbi- trageureven a 50/50 chance of cover- ing costs, or any compensation for

for routine cases.2 For ex-

1. Footnotes appear at end of article.

risk. Moreover, it is clear that, given the cost and uncertaintyof the trade, arbitrageurswill not take unlimited positions, even at pricesoutside these wide bounds. This makes for some- thing less than impenetrablebarriers. Weareleft with distressinglylittlein the way of a response to the skeptic's question.

Trading Real Options Using Model Prices The problems associated with apply- ing a theoretical option valuation model to the real world have obvi- ously not prevented virtually every serious option traderfrom doing just that. But they use the model in ways that are more consistent with the sec- ond wise man's approach than with the model's theoretical underpin- nings. Few investorsuse the model mainly for finding mispricedoptions that can be arbitragedagainst the underlying stock. A majorreason for this is that no one feels confident they know the true volatility. Estimating volatility from a sample of past prices is a rou- tine exercise,but you can'tbe sure the future will be exactly like the past. Basedon historicalvolatility,forexam- ple, the pricechange that occurredon October 19, 1987 was essentially


Option traders normally pay more attention to the impliedvolatility that sets the model option value equal to the marketprice. This is easy to com- pute and, in principle, gives a direct reading of the market'svolatilityesti- mate. But it has the disadvantagethat you have to assume the market is pricing the option according to the model, which rules out using implied

volatilityto detect mispricing. Another problem is that, because volatilityis the one input to the model that can't be directly observed, im- plied volatilityactuallyserves as a free parameter.It impounds expected vol- atility and everythingelse that affects option supply and demand but is not

in the model. Ifa changein the taxlaw

makes writing options less attractive, for instance, the priceeffectwill show

up in implied volatility. Any time the marketprices options differentlyfrom

a given model, for any reason, the

implied volatility derived from that model is not going to be the market's true volatilityestimate. Generally,implied volatilities differ across strike prices in a regularway, even though it is logicallyinconsistent for a stock to have more than one volatility. Out-of-the-money options typicallyhave higher implied volatili- ties than at and in-the-moneyoptions, puts are often priced on differentvol- atilities than calls, and the patterns vary from time to time. This is evi- dence that the marketdoes not price options strictly according to the model, or at least not accordingto the particularmodel thatproducesthe dif- fering implied volatilities. Tradersdon't caremuchaboutthese problems. In fact, they seldom think about the technicaldetails of the com- puterized models they use, or even about whether they are pricing Euro- pean or American options. Instead, they tend to take whatever theoretical model happens to be availableon the computerand treatthe impliedvolatil- ity it produces for a particularoption as a kind of index of how the option is currentlybeing priced in the market relativeto otheroptions and relativeto how it was priced at other times. It is perfectly normal, for example, for an option traderto reason, "Yes- terday the computer said these XYZ options were tradingon a volatilityof 20 per cent, but this morning the op- tions marketis a little soft, so I'll use 19.5 today and 21.5 for the out- of-the-money puts that always are a little richer." Thesevolatility"estimates"arethen plugged into the computer's model, and bids and offers are based on the theoreticalvalues thatit produces.The idea behind this procedureis not that the traderthinks 19.5 is the best avail- able estimate of XYZ'svolatility until option expiration,but that it is a rea- sonableway to summarizethe current state of supply and demand for XYZ options and that conditions will prob- ably remain fairlystable for a while- until they change. Formany option traders,having ex- actly the right model and volatility may not be of such great importance, because what they really careabout is the delta,which tells how to hedge an


option, and delta is much less sensi- tive to these things than theoretical value. Also, market-makersand active traders frequently hedge options againsteach other, ratherthan against the stock, so the effects of changing volatilityand othermodel inaccuracies on the different options partly offset each other. In other words, it may not matter so much if your model mis- prices an option as long as the option will continue to be mispriced in the same way when the stock price changes, or as long as it is hedged by other options that are similarly mis- priced. Still, when a stock has a large price move, implied volatilities also typi- cally change, meaning that the actual change in the option's price is not what was predicted by the original delta. The skepticmight wonder if we can even be sure that the apparent stability of implied volatility under normal circumstancesisn't partly due

to a kind

what with so many option tradersus- ing the model to change theirbids and offersas the stock price moves. While it was conceived as an arbi- trage-based valuation theory, the Black-Scholesmodel is actually used by option traders as a pricing equa- tion, to predict how the option price

will changewhen the underlyingstock moves. But traders treat the model almost as if it were a rule of thumb, rather than a formula that gives the true option value with confidence.

of self-fulfilling prophecy,

Testing Models with Real Option Prices

In a sense, the mirror image of an option traderevaluatingmarketprices using the model is the academictheo- rist "testing"a model on marketdata. The standardprocedureis to compute theoretical values for a set of actual options and compare them with mar- ket prices. Small random differences can be explained as "noise," but sys- tematic deviations are viewed as evi- dence of problems with the model. The common finding that deep-out- of-the-money options seem to be priced higher in the market than Black-Scholes would suggest has given rise to any number of increas-

ingly complex alternative models


explain the phenomenon. Tinkering with a model to try


make it fit market prices better con- fuses valuation and pricing. What an arbitrage-basedoption model says is that buying the option at its fairvalue and following the arbitragestrategy until expiration will return the risk- free rate of interest. If empiricaltests show that this would happen in prac- tice, assumingyou could buy the op- tion for its model value and had no

transactioncosts, then the valuation formulais correct.How those options might be priced in the market has nothing to do with whether the model valuesthem correctly! Whatmodel-testershave in mind, of course,is thatthe theoreticalvaluation model should also be the market's pricingequation. As we have seen, in the real world an arbitrage-basedval- uationmodel producesa band around the theoretical option value, within which the excess profit that could be made is smallerthan the cost to set up the arbitragetrade.In this context, the model only says that the price should be within the bounds. It is perfectly consistentwith the model forthe mar- ket price to be always at the lower bound or the upper bound, or follow- ing any kind of pattern in between. The finding of a consistent "bias"in marketpricesdoes not indicatea rejec- tion of the model, unless the bias is large enough that prices lie well out- side the arbitragebounds. And those bounds may be very wide.

Is the Skeptic Right?

A true skeptic might argue that this discussion has shown two things. First, option valuation models don't give correct fair values because the continuous arbitragecan't be done in practice. Second, even if models did give correct values, option prices in the marketwould not equal those val- ues because of all of the other factors affectingsupply and demand. I would not go so far, although I think some skepticism is a healthy, and risk-averse,attitudein this case. A model does not have to be exactly right for it to be of use. The general acceptanceof optionmodels in the real world, and the fact that extensive em-

pirical examination of those models has not led to widespread rejectionof them, suggests that, at least for short- maturity,exchange-tradedoption con- tracts, the models work acceptably well. But some situations warrant a greaterdegree of skepticismthan oth- ers. Thereare problemsof two kinds:

The model can be wrong, so that the theoreticalvalue is not the true option value, or the model may give the cor- rect value, but the market price the option differently.

In the first category we include


uations in which the model assump- tions are violated (to a larger extent than normal). This is true for long- maturity options, where parameters such as volatility and interest rates (which one can treat as being fairly constant over a month) can vary widely and unpredictablyover longer periods. We can have confidence in a

valuationmodel only to the extentthat we are confidentof our forecastsof its parametersout to option expiration, which may be many years in the fu- ture. We can also expectinaccuraciesaris- ing from errors in modeling security prices as geometricbrownian motion. There is considerable evidence that actualpricechanges have "fattails"-

that is, there is a greaterprobabilityof a largechange in a short intervalthan the model assumes, leading to hedg-


short-maturityoptions. There is also growing evidence that over both long and shorthorizons, thereis some non- randomnessin stockpricemovements.4 Thereis certainlyreasonto doubtan arbitrage-basedmodel's valuation of an option on an underlying asset that is not traded, even though theorists routinely apply the Black-Scholesfor- mula to all manner of cases in which the arbitrageis impossible, or essen- tially so. For example, it is an impor- tant theoreticalinsight that limited li- abilityin bankruptcymakesthe equity of a firmwith outstandingdebt similar to a calloption to buy the firm'sassets from the bondholders by paying the debt. But the firmas a whole is not an asset thatcanbe tradedindependently from its securities, so there is no way investors could arbitrage the stock

problems and undervaluation of


0 14

against the underlying firm if they thought the marketwas mispricingits option value. The same problem applies to op- tions on any nontraded asset, or on anything that is not an asset at all, such as the "option" to abandon an investment project. Even valuing an option on a marketablebut indivisible asset, such as a unique piece of prop- erty, causes difficultywhen thereis no way to forma hedge portfoliothat can be rebalanced.Thereare clearlymany cases in which one must be skeptical about deriving an option's value from an arbitrage strategy that cannot be done. The second category of problems pertains when the model gives the truevalue of an option, but the market pricesit differently.The more difficult and costly the arbitragetradeis to do, the greaterthe scope is for factorsnot covered in the model to move the marketprice away from its theoretical value. Several situations have proved particularly difficult for arbitrage- based models. Out-of-the-moneyoptions:People par- ticularlylike the combinationof a large potential payoff and limited risk and are willing to pay a premium for it. Thatis why they buy lottery ticketsat prices that embody an expected loss. Out-of-the-moneyoptions offera sim-

ilar payoff pattern. At the

the writers of those options are ex- posed to substanial risk because it is

hard to hedge against large price changes. Why should we not expect out-of-the-moneyoptions to sell for a premium over fairvalue? Americanoptions:The possibility of earlyexercisemakesAmericanoptions hard to value theoretically,especially because the early-exerciseprovisionis seldom exercised optimally according to the theory. This is an enormous problemwith mortgage-backedsecuri- ties because of the homeowner's op-

same time,

tion to prepay the mortgageloan, but all Americanoptions share it to some extent. We should not be surprisedif the market prices American options differentlyfromtheirmodel values be- cause of the uncertainty. Embeddedoptions:Valuationmodels treata securitywith embedded option features, such as a callablebond or a

securitywith defaultrisk, as if it were simply the sum of a straightsecurity and the option. But the market does not generally price things this way. For example, when coupon strippers unbundle government bonds, or when mortgagepass-throughsare re-

packaged into

parts sells for more than the original

whole. Why should we expect the marketto priceembeddedoptionsas if they could be tradedseparatelywhen this is not true of other securities? Timesofcrisis:Theperiodaroundthe crash of October1987showed that in times of financialcrisis, arbitragebe- comes even harder to do and option prices can be subject to tremendous pressures. At such times, we should not expectto be able to explainmarket prices well with an arbitrage-based valuationmodel.

CMOs, the sum of the

Where Do We Go FromHere?

If what is really wanted is a model to explain how the market prices op- tions, it doesn't make sense for aca- demics and buildersof option models to restrict their attention entirely to elaborating arbitrage-basedvaluation models in an ideal market. They should at least examine broader classes of theories that include factors such as expectations,riskaversionand market "imperfections"that do not enter arbitrage-basedvaluation mod- els but do affect option demand and supply in the real world. Forthose who would use theoretical models to trade actual options, it is safer to use models for hedging than

for computing option values; further- more, the harderthe arbitrageis to do, the less confidencethese investorscan have that the model is going to give either the true option value or the market price. Hedging options with options, ratherthan with the underly- ing stock, can provide some defense against inaccuratevolatility estimates and model misspecification. In general, investors who are not tradingas market-makingarbitrageurs should be less concerned with valua- tion models thanwith using options to produce overall payoff patterns that suit theirmarketexpectationsand risk preferences.Whenthey thinkthe mar- ket mightdrop sharply,it makessense for them to buy put options, even if they have to pay more than "fair" value.


1. See M. Rubinstein,"TheValuation of Uncertain Income Streams and the Pricingof Options," BellJournal ofEconomicsandManagementScience, Autumn 1976, or M. J. Brennan, "ThePricingof Contingent Claims in DiscreteTimeModels," TheJour- nalof Finance,March1979.

2. S. Figlewski,"OptionsArbitragein ImperfectMarkets," TheJournalof


3. And including that day's price

change in volatility estimates after the event meant that it dominated


calculation. There was then a

spurioussharpfallin estimatedvol- atilitymonths later, on the day Oc-

tober 19 dropped out of the data sample.

4. See, for example, E. Fama and K. French, "Permanent and Tempo- raryComponents of Stock Prices," Journalof PoliticalEconomy,April 1988,orJ.Poterbaand L. Summers, "MeanReversionin Stock Prices," Journalof FinancialEconomics,Octo- ber 1988.