option smile and the sabr model of stochastic volatility · (b) smile (or skew): for a given...
TRANSCRIPT
Options MarketsOptions Models
The SABR modelSome uses of SABR
Option Smile and the SABR Modelof Stochastic Volatility
Andrew Lesniewski
Baruch CollegeNew York
MITMarch 20, 2014
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Outline
1 Options Markets
2 Options Models
3 The SABR model
4 Some uses of SABR
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
LIBOR rate
LIBOR spot rate is a key benchmark interest rate for capital markettransactions:
Interest rate on a 3 month deposit, set daily in LondonUsed as index rate for setting cash flows on floating rateinstruments
OIS rate:
Based the Federal Funds effective rate, set dailyUsed for discounting cash flows
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
FRAs and swaps
FRAs:
Counterparty A agrees to pay counterparty B interest rateF on a 3 month deposit starting T years from nowThe break-even rate on a FRA is called the LIBOR forwardrateTraded over the counter (OTC)
Swaps:
Multi-period versions of FRAsThe break-even rate on a FRA is called the swap rateTraded OTC
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Options on LIBOR
The simplest options on a LIBOR forward rate are caps and floors.
A cap struck at K (say 2%) is the option to receive T years from nowthe interest
(F − K )+,
on a pre-specified notional principal. Here, x+ = max(x , 0).
Likewise, a floor struck at K is the option to receive T years from nowthe interest
(K − F )+.
Other commonly traded interest rate options are options on swaps a.k.a.swaptions. Swaptions can be viewed as options on forward swap rates.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Black’s model
The market standard for quoting option prices is in terms of Black’s model.We assume that a forward rate F (t), such as a LIBOR forward or a forwardswap rate, follows a driftless lognormal process reminiscent of the basicBlack-Scholes model,
dF (t) = σF (t) dW (t) .
Here W (t) is a Wiener process, and σ is the lognormal volatility. It isunderstood here, that we have chosen a numeraire Z with the property that,in the units of that numeraire, F (t) is a tradable asset. The process F (t) isthus a martingale, and we let Q denote the probability distribution.The solution to this stochastic differential equation reads:
F (t) = F0 exp(σW (t)− 1
2σ2t).
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Black’s modelConsider a European call struck at K and expiring at T . Its value today is
Call(T ,K ,F0, σ) = Z0E[(F (T )− K )+],
where E denotes expected value with respect to Q, and where Z0 is thecurrent value of Z. Explicitly,
Call(T ,K ,F0, σ) = Z0[F0N(d+)− KN(d−)
]Here,
N(x) =1√2π
∫ x
−∞e−y2/2dy
is the cumulative normal distribution, and
d± =log
F0
K± 1
2σ2T
σ√
T.
Similarly, the price of a European put is given by:
Put(T ,K ,F0, σ) = Z0[− F0N(−d+) + KN(−d−)
].
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Calibrating Black’s model
Black’s model is calibrated to the market if
Price (T ,K ,F0, σ) = MktPrice.
This equation can be solved for σ :
σimpl = ImplVol (T ,K ,F0,MktPrice) .
σimpl is known as the implied volatility.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Issues with Black’s model
The basic premise of Black’s model, that σ is independent of T , K , and F0, isnot supported by the markets. In fact, implied volatilities exhibit:
(a) Term structure: At the money volatility depends on the option expiration.
(b) Smile (or skew): For a given expiration, there is a pronounceddependence of implied volatilities on the option strike.
In order to accurately value and risk manage options portfolios, refinementsto Black’s model are necessary. Modeling term structure of volatility is hard,and not much progress has been made. We will focus on modeling volatilitysmile.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Volatility smile
0.6 0.8 1 1.2 1.4 1.6 1.8 279
80
81
82
83
84
85
86
87
Forward rate (%)
Impl
ied
vola
tility
(%)
MarketSABR
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Local volatility models
An extension of Black’s model is a class of models called local volatilitymodels.
The idea is that even though the exact nature of volatility of theunderlying asset is unknown, one can use an effective (“local”)specification of the underlying process so that the implied volatilitiesmatch the market implied volatilities.
Local volatility models are specified in the form
dF (t) = C(t ,F (t))dW (t) ,
where C(t ,F ) is an effective instantaneous volatility. It is convenient towork with a parametric specification of C(t ,F (t)) that fits the marketdata.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Local volatility models
Popular local volatility models which admit analytic solutions include:
(a) Normal model
(b) Shifted lognormal model
(c) CEV model
Option price in a local volatility model has the form
Price(T ,K ,F0, σ) = Z0B(T ,K ,F0, σ),
where B(T ,K ,F0, σ) = E[(F (T )− K )+].
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Normal modelThe dynamics of the forward F (t) in the normal model reads
dF (t) = σdW (t) .
The parameter σ is called the normal volatility. This is easy to solve:
F (t) = F0 + σW (t) .
This solution exhibits a drawbacks of the normal model: F (t) may becomenegative in finite time.The of a European call is given by:
Call(T ,K ,F0, σ) = Z0σ√
T(
d+N (d+) + N ′ (d+)),
whered+ =
F0 − Kσ√
T.
Many traders think in terms of normal implied volatilities, as the normal modeloften seems to capture the rates dynamics better than the lognormal model.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Shifted lognormal model
The shifted lognormal model is a diffusion process whose volatility structureis a linear interpolation between the normal and lognormal volatilities:
dF (t) = (σ1F (t) + σ0) dW (t) .
The volatility structure of the shifted lognormal model is given by the valuesof the parameters σ1 and σ0.The price of a call is given by the following valuation formula:
Call(T ,K ,F0, σ0, σ1) = Z0
((F0 + σ0/σ1)N(d+)− (K + σ0/σ1)N(d−)
),
where
d± =log
σ1F0 + σ0
σ1K + σ0± 1
2σ2
1T
σ1√
T.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
CEV modelThe CEV model, whose volatility structure is a power interpolation betweenthe normal and lognormal volatilities, is given by
dF (t) = σF (t)β dW (t) ,
where −∞ < β < 1. In order for the dynamics to be well defined, we specifya boundary condition at F = 0:
(a) Dirichlet (absorbing): F |0 = 0. Solution exists for all values of β, or
(b) Neumann (reflecting): F ′|0 = 0. Solution exists for 12 ≤ β < 1.
The CEV model requires solving a terminal value problem for the backwardKolmogorov equation:
∂
∂tB(t , x) +
12σ2x2β ∂2
∂x2 B(t , x) = 0,
B(T , x) =
{(x − K )+, for a call,(K − x)+, for a put,
with appropriate boundary condition at zero x .A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
CEV model
Pricing formulas for the CEV model can be obtained in a closed (albeitsomewhat complicated) form.For example, the Dirichlet call price can be expressed in terms of the thecumulative function Φ(x ; r , λ) of the non-central χ2 distribution as follows:
Call(T ,K ,F0, σ)
= Z0
(F0
(1− Φ
(ν2K 2/ν
σ2T ; ν + 2, ν2F2/ν
0σ2T
))− K Φ
( ν2F2/ν0
σ2T ; ν, ν2K 2/ν
σ2T
))where
ν =1
1− β , i.e.ν ≥ 1 .
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
SABR model
The volatility skew models that we have discussed so far improve on Black’smodels but still fail to reflect the market dynamics. One issue is, for example,the “wing effect” exhibited by the implied volatilities of some expirations(especially shorter dated) which is not captured by these models: the impliedvolatilities tend to rise for high strikes forming the familiar “smile” shape.Among the attempts to move beyond the locality framework are:
(a) Stochastic volatility models. In this approach, we add a new stochasticfactor to the dynamics by assuming that a suitable volatility parameteritself follows a stochastic process.
(b) Jump diffusion models. These models use a broader class of stochasticprocesses (for example, Levy processes) to drive the dynamics of theunderlying asset. These more general processes allow fordiscontinuities (“jumps”) in the asset dynamics.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
SABR modelThe SABR model is an extension of the CEV model in which the volatilityparameter follows a stochastic process:
dF (t) = σ (t) F (t)β dW (t) ,
dσ (t) = ασ (t) dZ (t) .
Here F (t) is a process which, may denote a LIBOR forward or a forwardswap rate, and σ (t) is the stochastic volatility parameter. The two Brownianmotions, W (t) and Z (t) are correlated:
E [dW (t) dZ (t)] = ρdt ,
where ρ is assumed constant. The parameter α is the volatility of σ (t) (thevolvol), and is also assumed constant. The dynamics is subject to the initialcondition:
F (0) = F0,
σ (0) = σ0,
where F0 is the current value of the forward, and σ0 is the current value of thevolatility parameter.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
SABR model
As in the case of the CEV model, the analysis of the SABR model requiressolving the terminal value problem for the backward Kolmogorov equation.Namely, the valuation function B(t , x , y) is the solution to
∂
∂tB +
12σ2(
x2β ∂2
∂x2 + 2αρxβ∂2
∂x∂y+ α2 ∂2
∂y2
)B = 0,
B(T , x , y) =
{(x − K )+ , for a call,(K − x)+ , for a put.
This is a more difficult problem than the models discussed above. Except forthe special case of β = 0, no explicit solution to this model is known. Thegeneral case can be solved approximately by means of a perturbationexpansion in the parameter ε = Tα2, where T is the maturity of the option.Luckily, this parameter is typically small and the approximate solution isactually quite accurate. Also significantly, this solution is very easy toimplement in computer code, and lends itself well to risk management oflarge portfolios of options in real time.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
The SABR model
There is no known closed form option valuation formula in the SABR model.Instead, one takes the following approach. We force the valuation formula inthe SABR model to be of the form
Call(T ,K ,F0, σ) = Z0[F0N(d+)− KN(d−)
],
Put(T ,K ,F0, σ) = Z0[− F0N(−d+) + KN(−d−)
],
d± =log
F0
K± 1
2σ2T
σ√
T,
given by Black’s model, with the implied volatility σ now a function of theSABR model parameters and the market data. This can be done by means ofan asymptotic expansion.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
SABR modelOne can show shows that the implied volatility is then approximately given by:
σ(T ,K ,F0, σ0,α, β, ρ) = αlog(F0/K )
D(ζ)
{1 +
[2γ2 − γ21 + 1/F 2
mid
24
×(σ0Fβmid
α
)2+ργ1
4σ0Fβmid
α+
2− 3ρ2
24
]ε+ . . .
},
where Fmid denotes a midpoint between F0 and K (such as (F0 + K )/2), and
γ1 =β
Fmid, γ2 =
β(β − 1)
F 2mid
.
The “distance function” entering the formula above is given by:
δ(ζ) = log(√1− 2ρζ + ζ2 + ζ − ρ
1− ρ
),
where
ζ =α
σ0
∫ F0
K
dxxβ
=α
σ0(1− β)(F 1−β
0 − K 1−β).
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Calibration of SABR
For each option maturity we need four model parameters: σ0, α, β, ρ. Wechoose them so that the model matches the market implied vols for severaldifferent strikes.
It turns out that there is a bit of redundancy between the parameters βand ρ. As a result, one usually calibrates the model by fixing β.
For a fixed β ≤ 1, say β = 0.5, we calibrate σ0, α, ρ. This works quitewell under “normal” conditions.
The SABR model can be calibrated to match the market in a variety ofconditions.
Calibration results show a persistent term structure of the modelparameters as functions of option expiration: typical is the shape of theparameter α which starts out high for short dated options and thendeclines monotonically as the option expiration increases.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Calibration of SABR
Alpha as function of option maturity
0.00
0.20
0.40
0.60
0.80
1.00
1.20
0 1 2 3 4 5 6
Maturity (years)
Alph
a
Figure: Historical values of the calibrated parameter ρ (β = 0.5)A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Calibration of SABR
Rho as function of option maturity
-0.40
-0.35
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
0.000 1 2 3 4 5 6
Maturity (years)
Rho
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Calibration of SABR
Generally, the model behaves very well under various market conditions.However:
In times of distress, such as 2008 - 2009, the height of the recentfinancial crisis, the choice of β = 0.5 occasionally led to extremecalibrations of the correlation parameters (ρ = ±1). As a result, somepractitioners choose high β’s, β ≈ 1 for short expiry options and let itdecay as option expiries move out.
On a few days at the height of the recent financial crisis the value of αcorresponding to 1 month into 1 year swaptions was as high as 4.7.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Calibration of SABR
Figure: Historical values of the calibrated parameter α (β = 0.5)
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Calibration of SABR
Figure: Historical values of the calibrated parameter ρ (β = 0.5)
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Arbitrage in the SABR model
The explicit implied volatility formulas make the SABR model easy toimplement, calibrate, and use. These implied volatility formulas are usuallytreated as if they were exact, even though they are derived from anasymptotic expansion which requires that α2T � 1. The unstated argumentis that instead of treating these formulas as an accurate approximation to theSABR model, they could be regarded as the exact solution to some othermodel which is well approximated by the SABR model. This is a validviewpoint as long as the option prices obtained using the explicit formulas forσ are arbitrage free. There are two key requirements for arbitrage freeness ofa volatility smile model:
(i) Put-call parity, which holds automatically since we are using the sameimplied volatility σ for both calls and puts.
(ii) The terminal probability density function implied by the call and putprices needs to be positive.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Arbitrage in the SABR modelTo explore the second condition, note that call and put prices can be writtenquite generally as
Call(T ,K ) = Z0
∫ ∞−∞
(F − K )+ qT (F ,K )dF ,
Put(T ,K ) = Z0
∫ ∞−∞
(K − F )+ qT (F ,K )dF ,
where qT (F ,K ) is the density of the risk-neutral probability Q at the exercisedate. Since
d2
dx2 x+ = δ(x),
it follows that
∂2
∂K 2 Call(T ,K ) =∂2
∂K 2 Put(T ,K )
= Z0qT (F ,K ) ≥ 0,
for all K .A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Arbitrage in the SABR model
The explicit implied volatility formula can represent an arbitrage free modelonly if
∂2
∂K 2 Call(T ,K ,F0, σ(K , . . .)) =∂2
∂K 2 Put(T ,K ,F0, σ(K , . . .))
≥ 0.
In other words, there cannot be a “butterfly arbitrage”.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Arbitrage in the SABR model
As it turns out:
It is possible for this requirement to be violated for very low strike andvery long expiry options.
The problem does not appear to be the quality of the option pricesobtained from the explicit implied volatility formulas, because these arequite accurate. Rather, the problem seems to be that implied volatilitycurves are not a robust representation of option prices for low strikes.
It is very easy to find a reasonable looking volatility curve σ(K , . . .)which violates the arbitrage free constraint for a range of values of K .
Implied volatility calculations in the SABR model can be done in a waythat guarantees arbitrage freeness.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Uses of SABR
Portfolio valuation (pricing)
Risk management
Relative value
Clearing and systemic risk management
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Hedging under SABR
The choice of volatility model impacts not only the prices of (out of themoney) options but also, at least equally significantly, their risk sensitivities.The issues one faces (among many others) are:
How is the delta risk defined: which volatility parameter should be keptconstant when taking the derivative with respect to the underlying?
How is the vega risk defined?
An answer to these questions may have a profound impact on the portfolioP&L.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Hedging under SABR
The delta risk of an option is calculated by shifting the current value of theunderlying while keeping the current value of implied volatility σ0 fixed. In thecase of an interest rate option, this amounts to shifting the relevant forwardrate without changing the implied volatility:
F0 → F0 + ∆F0,
σ0 → σ0,
where ∆F0 is, say, −1 bp. This scenario leads to the option delta:
∆ =∂Price∂F0
+∂Price∂σ
∂σ
∂F0.
The first term on the right hand side in the formula above is the original Blackmodel delta, and the second arises from the systematic change in the impliedvol as the underlying changes. This formula shows that both the classic deltaand vega contribute to the smile adjusted delta of an option.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Hedging under SABR
Similarly, the vega risk is calculated from
F0 → F0,
σ0 → σ0 + ∆σ0,
to beΛ =
∂V∂σ
∂σ
∂σ0.
These formulas are the classic SABR risk sensitivities (“greeks”).
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Hedging under SABRModified SABR greeks below do a better justice to the model dynamics.Since the processes for σ and F are correlated, whenever F changes, onaverage σ changes as well. This change is proportional to the correlationcoefficient ρ between the Brownian motions driving F and σ. It is easy to seethat a scenario consistent with the dynamics is of the form
F0 → F0 + ∆F0,
σ0 → σ0 + δFσ0.
HereδFσ0 =
ρα
Fβ0∆F0
is the average change in σ0 caused by the change in the underlying forward.The new delta risk is given by
∆ =∂V∂F0
+∂V∂σ
( ∂σ∂F0
+∂σ
∂σ0
ρα
Fβ0
).
This risk incorporates the average change in volatility caused by the changein the underlying.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Hedging under SABR
Similarly, the vega risk should be calculated from the scenario:
F0 → F0 + δσF0,
σ0 → σ0 + ∆σ0,
where
δσF0 =ρFβ0α
∆σ0
is the average change in F0 caused by the change in the beta vol. This leadsto the modified vega risk
Λ =∂V∂σ
∂σ
∂σ0+(∂V∂σ
∂σ
∂F0+∂V∂F0
)ρFβ0α
.
The first term on the right hand side of the formula above is the classic SABRvega, while the second term accounts for the change in volatility caused bythe move in the underlying forward rate.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Hedging under SABR
Figure: Regression of δFσ0 against ρα/Fβ for the 1Y into 10Yswaption β = 0.5.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
Hedging under SABRThe graph below shows the historical values of the calibrated parameter ρ.
Figure: Regression of δFσ0 against ρα/Fβ for the 5Y into 5Y swaptionβ = 0.75.
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility
Options MarketsOptions Models
The SABR modelSome uses of SABR
References
Agarwal, N., and McWilliams, G.: Evolution of volatility surface underSABR model, Courant Institute (2010).
Bartlett, B.: Hedging under SABR model, Wilmott Magazine,July/August, 2 - 4 (2006).
Doust, P.: No-arbitrage SABR, J. Comp. Finance 15, 3 31 (2012).
Hagan, P. S., Kumar, D., Lesniewski, A., and Woodward, D. E.:Managing smile risk, Wilmott Magazine, September, 84 - 108 (2002).
Hagan, P. S., Kumar, D., Lesniewski, A., and Woodward, D. E.:Arbitrage-free SABR, Wilmott Magazine, January, 60 - 75 (2014).
Hagan, P., Lesniewski, A., and Woodward, D.: Probability distribution inthe SABR model of stochastic volatility, preprint (2005).
A. Lesniewski Option Smile and the SABR Model of Stochastic Volatility