Abstract
As standard option pricing models rest on restrictive assumptions. additional tools which mitigate some of their deficiencies and facilitate their application have been developed. The main tools are based on the notion of implied volatility and on the sensitivities of the value of an option to the different model parameters. From a portfolio management perspective, it is useful to distinguish static from dynamic strategies. Section 12.1 presents the main static strategies, which consist in building an initial portfolio left unchanged until it is sold out. Section 12.2 examines the important notions of historical volatility, implied volatility, smile (or skew), and volatility surface. Section 12.3 analyzes the sensitivities of options to the different variables that influence their values. Section 12.4 studies dynamic strategies involving revisions and using these sensitivities as indicators for monitoring.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
In an alternative, more precise definition the initial investment is capitalized at T: G = V(T) – V(0)erT.
- 2.
Except at the points Ei and Kj since the functions Max(S–Ei, 0) and Max (Kj–S, 0) are not differentiable there.
- 3.
Recall that, in the BS model, between two dates i and i + 1 separated by a period of duration Δt (Δt = 1/p in what follows) we have \( \ln S\left(i+1\right)\hbox{--} \ln S(i)=\left(\mu \hbox{--} {\sigma}^2/2\right)\Delta t+\sigma \sqrt{\Delta t}U \), where U follows a reduced centered normal law. We use the notation \( {\mu}_p\equiv \left(\mu \hbox{--} {\sigma}^2/2\right)\Delta t\kern0.5em \mathrm{and}\kern0.5em {\sigma}_p\equiv \sigma \sqrt{\Delta t}=\sigma \frac{1}{\sqrt{p}} \).
- 4.
It is usual to express r and s as annual rates and T in fractions of a year. Any other choice of period is legitimate, if the dimensions of r, σ are T mutually coherent. σ is multiplied by \( \sqrt{p} \), since it is the variance σ2 which is proportional to time.
- 5.
We calculate “as if” no new information turns up on weekends or holidays.
- 6.
More exactly, as the more general model with stochastic interest rates shows, it is the average (between t and T) of the volatility of the underlying’s forward price which determines the standard deviation of the returns on ST and therefore the option price.
- 7.
Since the function F is continuous and monotonously increasing in σ.
- 8.
This type of (uniquely correct) reasoning is pushed to its logical limit on the official market for exchange rate options in Philadelphia (USA). There options are quoted not in dollars or any other currency, but directly in volatilities. Once an option has been traded on the basis of a volatility agreed by the two parties, the market’s back-office calculates the price, for example in dollars, using the Garman–Kohlhagen formula (see Chap. 11, Proposition 9).
- 9.
We are ignoring here the problem of dividends, the estimation of which may be difficult if the option is long-lived and its underlying asset pays several, usually random, dividends (which is the case for numerous warrants written on stocks).
- 10.
As shown in one of the appendices of the preceding chapter, N(d1) is exactly the exercise probability for the option under the measure QS associated to a special numeraire, namely the underlying itself: ProbQs(S(T) > K).
- 11.
If an option is in-the-money, \( {d}_1=\frac{r+{\sigma}_2/2}{\sigma}\sqrt{T} \) is near 0 for most of the current values of the parameters. More precisely: N(0) = 0.5, therefore \( N\left({d}_1\right)\sim N(0)+{d}_1{N}^{\prime }(0)=0.5+\frac{r+{\sigma}_2/2}{2\pi \sigma}\sqrt{T} \). The delta of a call in-the-money is therefore a little larger than 0.5, and that of a put a little larger than – 0.5.
N(d1) is even closer to this exercise probability for options called at-the-money-forward for which S(0) is not equal to the strike K but to the discounted strike Ke–rT (recall the call-put parity). Indeed, the rate r present in the preceding expressions then vanishes.
- 12.
Vega (the name of a star) is not a Greek letter, in contrast with the names of other common partial derivatives. At first it was called lambda (λ) or zeta (ζ), but the use of vega came into vogue with professionals. One continues for convenience to use generically the phrasing “Greek parameters” (even for cross second derivatives involving the volatility, such as the vanna or the voma).
- 13.
In contrast with the other Greeks, which are sensitivities to the model’s variables, the vega is a sensitivity to the model’s parameter.
- 14.
It is legitimate to apply the BS model in a context with variable rates, even stochastic rates; as we have seen in the preceding chapter, the relevant rate is the rate for maturity T and the relevant volatility is that of the forward price.
- 15.
We neglect the influence of terms “of order 2” (in fact, they are of order 1) except for (ΔS)2; the latter does have a homogeneous term in s2S2Δt (indeed ΔW2 = Δt). Other terms in the series expansion of order 2, also homogeneous in Δt ((Δs)2 and (Δs ΔS) ), are usually considered of lesser importance in practice.
- 16.
We do not take into account the variation in the possibly existing dividend rate.
- 17.
More precisely, we assume this risk is independently monitored by taking positions in zero-coupons matching the options’ maturities.
- 18.
As a rule, the initial value of a self-financing portfolio , V(0), is not zero and so brings with it a positive (negative) holding cost (gain) linked to the risk-free rate r over the agent’s investment horizon.
- 19.
What is relevant is the net P&L, as the carrying cost (gain) rV reduces (increases) θ so that (12.12) holds.
- 20.
The proposition does in fact hold more generally than it may seem and is also applicable to positions that are not delta-neutral but may have any (target) delta. The gamma is the number of shares it is required to buy or sell to recover the desired, target delta.
- 21.
That is always true for a position with zero value and, if not, it is still true if one takes the portfolio’s financing cost into account.
- 22.
If the option is near the money, its gamma increases and the parabola becomes sharper. Conversely, the gamma of an option far from the money decreases and the parabola becomes flatter.
- 23.
It is, in any case, important to note that, for market makers, overturning positions is constrained by their statutory obligation to serve the market. Recall that the role of market makers is to ensure the market’s liquidity: within the limits of the bid-ask spreads that they post, they must buy or sell a minimum quantity of options to any market participant who requests it.
- 24.
Precisely, one sells a number δ2 of options O1 and buys a number δ1 of options O2 so that –δ2δ1 + δ1δ2 = 0.
- 25.
To control for the delta, there has to be at least one option (in addition to the underlying); for every additional Greek to be controlled, there has to be at least one additional option.
- 26.
That is, the risk which is not generated by the variations in S and whose market price is not necessarily zero. See, for instance, Heston’s model (last Section of Chap. 11).
- 27.
Recall that for an asset not paying dividends: F0,T = S0erT . Furthermore, we have in all cases FT,T = ST.
- 28.
More precisely, this last equation written for x = ST shows the payoff Ψ(ST) is obtained using a static portfolio created on date 0 and containing: a zero-coupon yielding \( \Psi \left({F}_{0,T}\right)\hbox{--} {F}_{0,T}\frac{\partial \psi }{\partial k}\ \left({F}_{0,T}\right) \) on T; \( \frac{\partial \psi }{\partial k}\left({F}_{0,T}\right) \) units of the underlying; and a continuum \( \frac{\partial^2\Psi}{\partial {k}^2}(k) dk \)of options (call or put) out of the money and with strike k.
Suggestions for Further Reading
Books
Baxter, M., & Rennie, A. (1999). Financial Calculus. Press.
*Bergomi, L. (2016). Stochastic volatility modeling. Chapman & Hall.
*Fusai, G., & Roncoroni, A. (2008). Implementing models in quantitative finance: Methods and cases. Springer.
**Gatheral, J. (2006). The volatility surface: A practitioner’s guide. Wiley Finance.
Hull, J. (2018). Options, futures and other derivatives (10th ed.). Prentice Hall Pearson Education.
Kolb, R. (2007). Futures, options and swaps (5th ed.). Blackwell.
Taleb, N. (1997). Dynamic hedging: Managing vanilla and exotic options. Wiley.
Articles
Bakshi, G., Cao, C., & Chen, Z. (1997). Empirical performance of alternative option pricing models. Journal of Finance, 52(5), 2004–2049.
Bates, D. S. (2000). Post-’87 Crash fears in the S&P futures market. Journal of Econometrics, 94, 181–238.
Bergomi, L. (2004). Smile dynamics. Risk, 117–123.
Bergomi, L. (2005). Smile dynamics II. Risk, 67–73.
Bergomi, L. (2008). Smile dynamics III. Risk, 90–96.
Black, F. (1989). How to use the holes in Black-Scholes. Journal of Applied Corporate Finance, 1, 67–73.
Chaput, J. S., & Ederington, L. H. (2003). Option spread and combination trading. Journal of Derivatives, 10(4), 70–88.
Cookson, R. (1993). Moving in the right direction. Risk.
Daglish, T., Hull, J., & Suo, W. (2007). Volatility surfaces: Theory, rules of thumb, and empirical evidence. Quantitative Finance, 7(5), 507–524.
Derman, E. (1999). Regimes of volatility. Risk, 55–59.
Dumas, B., Flemming, J., & Whaley, R. (1998). Implied volatility functions: Empirical tests. Journal of Finance, 53-6, 2059–2106.
Figlewski, S. (1989). Options arbitrage in imperfect markets. Journal of Finance, 44, 1289–1311.
Harvey, C., & Whaley, R. (1991). S&P 100 Index option volatility. Journal of Finance, 46, 1551–1661.
Jackwerth, J., & Rubinstein, M. (1996). Recovering probability distributions from option prices. Journal of Finance, 51, 1611–1631.
Xu, X., & Taylor, S. J. (1994). The term structure of volatility implied by foreign exchange options. Journal of Financial and Quantitative Analysis, 29, 57–74.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1
1.1 Computing Partial Derivatives (Greeks)
To simplify notation, in this appendix we always consider that the current time is t = 0.
1.1.1 The Black–Scholes Model
Preliminary Result
Let us start with a result that simplifies calculating Greek parameters . Since the Black–Scholes model writes
with
we have
-
(i)
SN′(d1) = Ke–rTN′(d2)
where \( {N}^{\prime }(z)\equiv \frac{1}{\sqrt{2\pi }}{e}^{-{z}^2/2} \) is the density of the reduced centered normal distribution.
Proof
From the definition of d1:
which proves (i).
-
(a)
Delta
Moreover, from call-put parity we have P = C – S + Ke–rT, from which
-
(b)
Gamma
since \( N\left({d}_1\right)=N\left(\left(\ln \left(S/K\right)+\left(r+{\sigma}^2/2\right)T\right)/\sigma \sqrt{T}\right) \).
ΓP = ΓC from call-put parity (its second derivative).
-
(c)
Vega
from which
since \( {d}_1=\left(\ln \left(S/K\right)+r\ T\right)/\sigma \sqrt{T}+\sigma \sqrt{T}/2 \) and
and
Consequently, and since \( \frac{d_2}{\sigma }=\frac{d_1}{\sigma }-\sqrt{T} \), we have
from (i). Furthermore, \( {\upsilon}_P=\frac{\partial P}{\partial \sigma }={\upsilon}_C \), by call-put parity.
-
(d)
Theta
The direct proof gives us
One can, alternatively, recall that the option price satisfies the Black-Scholes partial differential equation in its form (12.10). Hence we deduce the following equation relating sensitivities
and therefore
Replacing the gamma and delta by the expressions given above, we recover the expression for theta.
Furthermore, again from call-put parity,
-
(e)
Rho
The rho of a put can then be obtained from call-put parity:
1.1.2 Other Models
We start from the Black-Scholes formula:
with \( {d}_1(S)=\left(\ln \left(S/K\right)+\left(r+{\sigma}^2/2\right)T\right)/\sigma \sqrt{T} \) and \( {d}_2(S)={d}_1(S)\hbox{--} \sigma \sqrt{T} \); moreover, we know BS′ = N(d1).
-
(a)
Black’s model (option on a forward contract quoted F)
The model gives the premium as a function of F: C = BS (Fe–rT) where the Black-Scholes BS(.) is defined above and \( {d}_1(F)=\left(\ln \left(F/K\right)+{\sigma}^2T/2\right)/\sigma \sqrt{T} \). From this, using the Greek parameters of the Black-Scholes model, we obtain almost immediately
\( {\delta}_C=\frac{\partial C}{\partial F}=\frac{\partial }{\partial F}\mathrm{BS}\left({\mathrm{Fe}}^{\hbox{--} \mathrm{rT}}\right)={\mathrm{e}}^{\hbox{--} \mathrm{rT}}{\mathrm{BS}}^{'}\left({\mathrm{Fe}}^{\hbox{--} \mathrm{rT}}\right) \), from which: δC = e-rTN(d1).
\( {\Gamma}_C=\frac{\partial^2 BS\left({Fe}^{- rT}\right)}{\partial {F}^2}={e}^{-2 rT}{BS}^{\prime \prime}\left({Fe}^{- rT}\right)={e}^{-2 rT}\frac{N^{\hbox{'}}\left({d}_1\right)}{F{e}^{- rT}\sigma \sqrt{T}} \), from which \( {\Gamma}_C={e}^{- rT}\frac{N^{\hbox{'}}\left({d}_1\right)}{F\sigma \sqrt{T}} \).
and
The Greek parameters related to a put can be obtained from those for the analogous call using the parity relationship written as P = C – e–rT (F – K), whence:
-
(b)
Merton’s Model (When the Underlying Distributes a Continuous Dividend c)
This model writes: \( C= BS\left({Se}^{\hbox{--} cT}\right);\kern0.5em {d}_1=\left(\ln \left(S/K\right)+\left(r\hbox{--} c+{\sigma}^2/2\right)T\right)/\sigma \sqrt{T} \) and \( {d}_2={d}_1\hbox{--} \sigma \sqrt{T} \).
From this, we have:
The Greek parameters for a put can be obtained from those for a call using the parity relation written as P = C – (Se–cT – Ke–rT).
Appendix 2
1.1 Option Prices and the Underlying Price Probability Distribution
The following proposition highlights the relationship between the prices of standard options and the probability distribution for the future value ST of the underlying.
Proposition
The probability density under Q of the final value ST of the underlying, \( {\phi}_T(k)=\frac{Q\left({S}_T\in \left[k,k+ dk\right]\right)}{dk} \), is equal to the capitalized value of the second derivative of the option (a call or a put) with respect to the strike denoted by k:
Proof
This result follows directly from the definition of the premium for calls or puts as the risk-neutral expectation of the discounted final payoff. As an example, take a call with strike k whose value at a distance T from its expiration date is
Taking the derivative of the right-hand side with respect to k, we have
Taking the derivative a second time, we obtain
An analogous proof can be used for the put.
This proposition shows that knowing the prices of vanilla options in theory allows us to deduce the distribution for the underlying’s future value implicitly used by the market for valuing options. This would be perfect knowledge in the purely theoretical case of a continuum of prices. In practice, of course, price discontinuity makes it imperfect and not so easy to use. In any event, it is theoretically possible to compare the distribution for ST which is implicit in option market prices with the theoretical distributions used in valuation models. In the special case of BS, the underlying’s distribution is assumed to be log-normal. In practice, the distribution deduced from vanilla option prices using Eq. (12.13) (T-density) seems to differ from the log-normal model. Empirical studies clearly show that, for stocks as underlyings, the left tail of the distribution for small values of ST is noticeably thicker than what follows from the Gaussian model. And this form of distribution for ST can explain why the smile is downward sloping (a skew , as in Fig. 12.12a), which is precisely the usual case for stock options. This connection between the shape of the distribution for ST and the smile or skew can be seen from Fig. 12.12a, b. Suppose the smile is decreasing like the solid line in Fig. 12.12a; also in Fig. 12.12a the horizontal smile assumed in the BS model is shown as a dotted line which serves as a reference. Assume further that the distribution of ln(ST) is asymmetrical, with a thick left tail, as drawn with a solid line in A-b ; the normal distribution (as hypothesized in BS) is also shown in the same figure, drawn dotted as a reference. The solid (dotted) lines in A-a thus correspond to the solid (dotted) lines in A-b.
On the abscissa axis K in Fig. 12.12a we find the underlying’s value S0 on date 0 (today). For an option with a low strike, such that K1 < S0, options are expensive in comparison to a standard BS price; the opposite is true for an option with a high strike, such that K2 > S0, which is relatively cheap.
Now let us consider a put deep out-of-the-money with strike K1; it is logical that it should be relatively costly (see A-a) since, as A-b shows, the probability of exercise (ST < K1) and the expected gain upon exercise are larger than for a log-normal distribution. It also is reasonable that a call deep out-of-the-money K2 should be cheap since the right tail of the distribution for ST is thinner than in the log-normal case. Finally, let us remark that, from the call-put parity, a relatively expensive put implies a relatively expensive call, and conversely, since C – P = S – Ke–rT, independently of any assumption regarding the actual distribution of ST.
The relative dearness of stock options with low strikes and the relative cheapness of options with high strikes (tantamount to a decreasing implied volatility curve) are therefore compatible with empirical results on the distributions of stock yields.
Similarly, a U-shaped smile curve (si (K) decreasing then increasing again) can be explained by an empirical distribution for ST with thick tails on both left and right.
This result shows that some assumptions of the Black–Scholes (BS) model are not realistic and, because of this, the model yields biased results. When applying BS formula, taking the smile into account corrects for this bias only imperfectly. The fundamental assumption that should be questioned is the constancy of the volatility parameter (which implies the log-normality of the underlying’s distribution).
An approach that is theoretically more rigorous and more general is to look for a model of which BS would be a special case, and which would be compatible with the whole volatility surface observed at a given instant (and not only with the smile curve relative to a specific expiration date T). Mathematically one has to find a model for the evolution of the underlying price which provides, for any future value of the underlying, a probability distribution equal to the distribution implied by the surface.
Several answers appear satisfactory and are used in practice. The first is to work with a model for the volatility s(t,S) that depends entirely on the time remaining until the option’s expiration date and the current price of the underlying. Such models can also explain the smile. It is, for example, in line with the empirical evidence and the theory that a stock’s volatility is a decreasing function s(S) of its price S. At the theoretical level, a drop in the stock price leads to an increase in the ratio (debt/market capitalization) (financial leverage); this increase in leverage implies an increase in the risk affecting the equity, and so an increased stock volatility. The fact that the stock volatility is higher for small than for large values of S explains the thick left tail of the distribution for ST (see Fig. 12.12b) and the volatility skew (see Fig. 12.12a).
A complementary explanation for this dual phenomenon of volatilities asymmetric to the right and returns asymmetric to the left is the feedback effect. Following a large positive jump in the price, the stock volatility increases which exerts pressure to reduce its return. Indeed, since the expected (ex ante) return must increase to offset the increased risk, the price must fall and with it the realized (ex post) return. This feedback phenomenon thus mitigates the amplitude of positive returns. In the same way, a large negative jump in the price causes an increase in volatility which forces the return down for the same reason as above. Thus the feedback effect accentuates the amplitude of negative returns. Consequently, this dissymmetric feedback effect contributes to the left asymmetry in the return distribution and the right asymmetry in the volatility distribution.
The interest of these models which express the volatility as a function s(t,S) lies in their relative simplicity; in particular, they can be implemented using Cox-Ross-Rubinstein recombining trees. An alternative solution consists of models with stochastic volatility, in which the variable st is itself the solution to a stochastic differential equation. Such models present the advantage to take into account the specific risk of a change in the volatility.Footnote 26
Appendix 3
1.1 Replication of an Arbitrary Payoff with a Static Option Portfolio
Let us consider European calls and puts with the same underlying and the same expiration date T. Let us denote by F0,T the forward priceFootnote 27 of the underlying on date 0 for an expiration date T. An option with strike k is said to be at-the-money-forward when F0,T = k. Subsequently, C(k) and P(k) respectively denote the premium at 0 of a call and a put with expiration date T and strike k. The following proposition describes the replication of an arbitrary payoff Ψ(ST) by a portfolio of calls and puts out-of-the-money-forward .
Proposition
Let us consider a security Ψ which generates a single payoff on date T defined by a sufficiently regular function Ψ(ST).
-
This security is worth on date 0:
-
The static duplication of the security Y is guaranteed by a position with \( \frac{\partial^2\Psi}{\partial {k}^2}(k) dk \) options (calls or puts) out-of-the-money and with strike k.
One says the vanilla calls and puts span the set of regular payoffs, for a given expiration date.
Proof
The proposition can be deduced from the following result: For any x > 0,
This equation can be justified with a bit of functional analysis. By definition of the first derivative, we have
Integrating by parts (consider u = x – k and \( v=\frac{\mathrm{\partial \Psi }}{\partial k}(k) \)), we find
But x − k = (x − k)+ − (k − x)+. Hence
Remark that for any x > 0, one may replace the integration limit x in the first integral by infinity (since, if k > x, (x – k)+ vanishes), and the limit x in the second integral by 0 (since, if k < x, (k – x)+ vanishes); as a result we get
Rewriting this, we have
This last equation shows that the replication of the payoff Y(ST) is ensured by a static position containing \( \frac{\partial^2\Psi}{\partial {k}^2}(k) dk \) options (call or put) out-of-the-money with strike k.Footnote 28
Expressing the discounted risk-neutral expectation of each member of the last equation for x = ST and noting, on the one hand, that EQ[(ST − Φ0, T)e−rT] = 0, and, on the other, that EQ[(ST − k)+e−rT] = C(k) and EQ[(k − ST)+e−rT] = P(k), we end up with Eq. (12.14).
The proposition proves that any payoff can be replicated by a static strategy based on vanilla call and put options. In a very intuitive sense, it is the gamma (second derivative of the payoff) on the expiration date that determines the quantity of standard options to hold in the replicating portfolio.
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this chapter
Cite this chapter
Poncet, P., Portait, R. (2022). Option Portfolio Strategies: Tools and Methods. In: Capital Market Finance. Springer Texts in Business and Economics. Springer, Cham. https://doi.org/10.1007/978-3-030-84600-8_12
Download citation
DOI: https://doi.org/10.1007/978-3-030-84600-8_12
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-84598-8
Online ISBN: 978-3-030-84600-8
eBook Packages: Economics and FinanceEconomics and Finance (R0)