Equivalent Black Volatilities¤

Patrick S. Hagany Diana E. Woodwardz

March 30, 1998

Abstract
We consider European calls and puts on an asset whose forward price F(t) obeys

dF (t) = ®(t)A(F )dW (t)

under the forward measure. By using singular perturbation techniques, we obtain explicit
algebraic formulas for the implied volatility ¾ B in terms of today’s forward price F0 ´ F (0),
the strike K of the option, and the time to expiry t ex . The price of any call or put can then be
calculated simply by substituting this implied volatility into Black’s formula.
For example, for a power law (constant elasticity of variance) model

dF (t) = aF ¯ dW (t)

we obtain
a n (1 ¡ ¯)(2 + ¯) ³ F0 ¡ K ´2 o
(1 ¡ ¯)2 a2 tex
¾B = 1¡¯
1+ + 2¡2¯
+ : ::
fav 24 fav 24 fav

where fav = 12 (F0 + K).

Our formula for the implied volatility is not exact. However, we show that the error is
1
insigni…cant, rarely approaching 1000 of the time value of the option. We also present more
accurate (albeit more complicated) formulas which can be used for the implied volatility.

¤The views presented in this report do not necessarily re‡ect the views of NumeriX, The Bank of Tokyo-Mitsubishi,

or any of their a¢liates.

y Current mailing address: NumeriX, 546 Fifth Avenue, 17th Floor, New York, NY 10036.
z Current mailing address: The Bank of Tokyo-Mitsubishi, Ltd., 1251 Avenue of the Americas, New York, NY

10020.

1
 Market prices of many European options are quoted in terms of their Black-Scholes volatility
¾ B . To obtain the cash price, let tex be the option’s exercise date, let ts be the option’s settlement
date, and de…ne F (t) to be the forward price of the asset (for a contract maturing at the settlement
date ts ) as seen at date t. Then the cash price of calls and puts is given by Black’s formula [1, 6]
© ª
(1.1a) V call = D(0; ts) F 0 (d1 ) ¡ K (d2 )
© ª
(1.1b) V put = D(0; ts) K (¡d2 ) ¡ F 0 (¡d1 )
with
log(F 0 =K ) § 12 ¾ 2B tex
(1.1c) d1;2 = p :
¾ B tex
Here F 0 ´ F (0) is today’s forward value of the asset, D(0; ts ) is today’s discount factor to the
settlement date, and ¾ B is the quoted volatility of the option.
The derivation of Black’s formula presumes that the forward prices F (t) are distributed log
normally about today’s forward price F 0 ,
(1.2) dF = ¾ B F dW; F (0) = F0 :

However, prices are rarely distributed log normally. Consequently market volatilities ¾ B usually
vary with the strike K and time-to-exercise tex .
Changing ¾ B with the strike and exercise date essentially means that a di¤erent model is being
used for each strike and expiry. This presents several di¢culties when managing large books of
options. First, it is not clear that the delta and vega risks calculated at a given strike are consistent
with the same risks calculated at other strikes, which interjects uncertainty into the consolidation
and hedging of risks across strikes. Second, if ¾ B varies with the strike K, it seems likely that ¾ B
also varies systematically as the forward price F changes [3, 4]. Any vega risk arising from the
systematic change of ¾ B with F could be hedged more properly (and inexpensively) as delta risk.
Finally, it is di¢cult to know which volatility to use in pricing exotic options.
An alternative approach is to use a single model which correctly prices options at di¤erent strikes
and exercise dates without “adjustment.” Commonly these models are of the form [3, 4]
(1.3) dF = ®(t)A(F )dW; F (0) = F 0

under the forward measure, with call and put prices given by the expected values
© ª
(1.4a) V call = D(0; ts )E [F (tex ) ¡ K]+ j F (0) = F0
© ª
(1.4b) Vput = D(0; ts )E [K ¡ F (tex )]+ j F (0) = F 0
in this measure.
In Appendix A we use singular perturbation techniques [9] to analyze these models and …nd
explicit expressions for the values of European calls and puts. These formulas are then used to
obtain the implied volatilities ¾ B of the options. For both calls and puts, we …nd that the implied
volatility is
ajA(f av )j n £ 00
1 A
¡ A 0 ¢2 2 ¤
¾B = 1 + 24 ¡2 + 2 (F0 ¡ K )2
fav A A fav
(1.5a) o
£ A 00 ¡ 0¢
A 2 1 ¤
1
+ 24 2 ¡ + 2 a 2A2 (fav )tex + : : : :
A A fav

1
Here A and its derivatives are to be evaluated midway between today’s forward price and the strike,
at

(1.5b) f av = 12 (F0 + K);

and a is the “sum-of-squares average” of ®(t),

Z tex
¡ 1 ¢1=2
(1.5c) a= ® 2 (t0)dt0 :
tex 0

For example consider a power law model

(1.6) dF = ®(t)F ¯ dW:

This model is just the constant elasticity of variance (CEV) model [2] written in terms of the forward
price F (t). For this model the implied volatility is

a n ³ F ¡ K ´2 a2 tex o
1 0 1
(1.7) ¾B = 1¡¯
1+ 24
(1 ¡ ¯)(2 + ¯) + 24
(1 ¡ ¯)2 2¡2¯
+ ::: ;
f av fav fav

where f av and a are given by (1.5b) and (1.5c) as before.

The “equivalent vol” formula (1.5) provides a very quick and very simple method for pricing
calls and puts. Instead of using lattice, PDE, or Monte Carlo methods to compute the value of the
option, one simply uses (1.5) to obtain the implied volatility ¾ B , and then substitutes this ¾ B into
Black’s formula (1.1).
Option hedges can also be obtained by writing (1.5) as

(1.8a) ¾ B ´ ¾ B (F0 ; K; tex)

and then writing the option price as

(1.8b) V = VB lack (F0 ; K; tex ; ¾ B (F 0 ; K; tex));

where V B lack is the call or put formula in (1.1). Di¤erentiating (1.8b) with respect to F 0 yields

@V @V Black @¾ B @V B lack
(1.9) = + :
@F 0 @F 0 @F0 @¾ B

Thus, the delta risk of the option is composed of two terms: the usual delta risk from Black’s formula,
plus a term proportional to the Black vega. This last term arises from the systematic change in the
implied volatility ¾ B of the option caused by changes in the forward price F .
The equivalent vol formula (1.5) is not exact. To test its accuracy, we compared the option
values obtained using the equivalent vol formula (1.5) against the “exact” option prices obtained by
numerical computation for di¤erent A(F ). We found that the equivalent vol formula is surprisingly
accurate, usually as accurate as an excellent tree or PDE valuation, and much more accurate than
Monte Carlo solutions. To demonstrate its accuracy, consider the power law model (1.6). For power
law models, we found that the worst errors occur when ¯ = 0, which is to be expected since ¯ = 0 is
“furthest” from the log normal Black model. Accordingly, in …gures 1-3 we show the error between
the exact option price and the price obtained using the equivalent vol formula (1.7) when ¯ = 0.
Shown is the error as a function of the strike K for 1 year, 2 year, 5 year, and 10 year options.
Figure 1 exhibits the error as the di¤erence between the exact implied vol and the equivalent vol

2
obtained from (1.7); …gure 2 exhibits the error as a fraction of the forward price F 0 ; and …gure 3
exhibits the error as a fraction of the option’s time value.
The singular perturbation analysis in Appendix A can be carried out to arbitrarily high order. In
Appendix B we state a more accurate equivalent vol formula obtained by carrying out the analysis
through O(²4 ). Figure 4 shows the error made using this more accurate formula. Although this
formula is clearly much more accurate than (1.5), we believe that (1.5) is precise enough to make
this improvement super‡uous in most cases.
Singular perturbation techniques can be used to solve many related pricing problems. For exam-
ple, these techniques can be used to solve intrinsically time-dependent models

(1.10) dF = A(t; F )dW

to obtain accurate equivalent vol formulas for calls and puts; to analyze one- and two-factor term-
structure models to obtain accurate equivalent vol formulas for swaptions, ‡oors, and caps; and to
obtain quick and accurate pricing of many exotics.

A Singular Perturbation Expansion

Consider a European call with expiration date tex , settlement date ts , and strike K . As before, let
F (t) be the stochastic process for the forward price as seen at date t. We are assuming that

(A.1) dF = ®(t)A(F )dW

under the forward measure. Under this measure, the value of the option at date t is V (t; F (t)),
where the function V (t; f ) is given by the expected value [7, 8]
© ª
(A.2) V (t; f ) = D(t; ts)E [F (tex ) ¡ K]+ j F (t) = f :

Here D(t; ts ) is the discount factor to the settlement date ts at date t.

To simplify the problem, let us strip the discount factor from V ,

(A.3) V (t; f ) ´ D(t; ts)Q(t; f ):

Then Q(t; f ) is de…ned by

© ª
(A.4a) Q(t; f ) = E [F (tex ) ¡ K ]+ j F (t) = f ;

and the expectation is over the probability distribution generated by the process

(A.4b) dF (t) = ®(t)A(F )dW (t):

Therefore Q(t; f ) satis…es the backward Kolmogorov equation [10]

(A.5a) Q t + 12 ®2 (t)A2 (f )Qf f = 0; t < tex

with the …nal condition

(A.5b) Q = [f ¡ K]+ at t = tex :

3
A.1 Scaling and asymptotic solution
To obtain the option value, we …rst scale equtions (A.5a) and (A.5b) appropriately, and then use
singular perturbation methods to solve the scaled problem. To obtain the equivalent vol, we analyze
Black’s model to determine the volatility ¾ B which would yield the same value of the option.
To scale the equations appropriately, de…ne

(A.6a) ² ´ A(K) << 1;

and the new variables

Z te
1 ~ ; x) = 1 Q(t; f ):
(A.6b) ¿ (t) = a2 (t0)dt0 ; x = (f ¡ K); Q(¿
t ² ²

In terms of the new variables, (A.5) becomes

~¿ ¡ 1 A2 (K + ²x) ~
(A.7a) Q 2
Qxx = 0 for ¿ > 0
A2 (K )

with

(A.7b) ~ = x+
Q at ¿ = 0:

~ ; x), the option value will be given by

After solving (A.7) for Q(¿
¡ ¢
(A.8) ~ ¿ (t); f ¡K :
V (t; f ) = D(t; ts )A(K) Q A(K )

By expanding
© ª
(A.9a) A(K + ²x) = A(K) 1 + ²º 1 x + 12 ²2 º 2 x2 + : : :

where

(A.9b) º 1 = A0(K)=A(K); º 2 = A00(K)=A(K); : :: ;

problem (A.7) becomes

(A.10a) ~¿ ¡ 1Q
Q ~ = ²º 1 xQ~ xx + 1 ² 2 (º 2 + º 2 )x 2 Q
~ xx + : : : ; ¿ >0
2 xx 2 1

with

(A.10b) ~ = x+
Q at ¿ = 0:

~ as
We can solve (A.10) by expanding Q

(A.11) ~ ; x) = Q0 (¿ ; x) + ²Q 1 (¿ ; x) + ²2 Q 2 (¿ ; x) + : : : :
Q(¿

Substituting (A.11) into (A.10) and equating like powers of ² yields a hierarchy of problems [9]. At
leading order, we obtain

(A.12a) Q0¿ ¡ 12 Q0xx = 0 for ¿ > 0;

(A.12b) Q0 = x+ at ¿ = 0:

4
At order ², we have

(A.13a) Q1¿ ¡ 12 Q1xx = º 1 xQ0xx for ¿ > 0;

(A.13b) Q1 = 0 at ¿ = 0:

And at order ² 2,

(A.14a) Q2¿ ¡ 12 Q 2xx = º 1 xQ 1xx + 12 (º 2 + º 21 )x2 Q 0xx for ¿ > 0;

(A.14b) Q2 = 0 at ¿ = 0;

and so on.
The solution of (A.12a) yields
¡ ¢ q ¿ ¡x2 =2¿
(A.15a) Q 0 (¿ ; x) ´ G(¿ ; x) = x px¿ + 2¼ e ;

as can be veri…ed by direct substitution. Note that G(¿ ; x) is essentially the value of a call option
using a normal (as opposed to a log normal) model for the forward price. For later convenience, we
note that
2
1 e¡ x =2¿
(A.15b) G¿ = 2
p ; Gx = ( px¿ );
2¼¿

2 2 2
x2 ¡ ¿ e¡ x =2¿ x e¡x =2¿ e¡ x =2¿
(A.15c) G¿ ¿ = p ; Gx¿ =¡ p ; Gxx = p ;
4¿ 2 2¼¿ 2¿ 2¼ ¿ 2¼¿

2
x4 ¡ 6x 2 ¿ + 3¿ 2 e¡ x =2¿
(A.15d) G¿ ¿ ¿ = p :
8¿ 4 2¼¿

At O(²), we substitute Q0 ´ G(¿ ; x) into (A.13) and use Gxx = 2G ¿ and (A.15c). This yields

(A.16a) Q1¿ ¡ 12 Q1xx = º 1 xG xx = ¡2º 1 ¿ Gx¿ for ¿ > 0

(A.16b) Q1 = 0 at ¿ = 0:

The solution of (A.16) is

(A.17) Q 1 = ¡º 1 ¿ 2 Gx¿ ´ º 1 ¿ xG¿ ;

as can be veri…ed by direct substitution.

At O(²2 ), we substitute Q 0 = G and Q1 = º 1¿ xG ¿ into (A.14). This yields
³ x4 ¡ 2¿ x2 ´ e¡x2 =2¿
(A.18a) Q 2¿ ¡ 12 Q 2xx = º 21 + 12 º 2 x2 p for ¿ > 0
2¿ 2¼¿

5
(A.18b) Q2 = 0 at ¿ = 0:

The solution of (A.18) is

£ ¤ £ ¤
(A.19) Q 2 = º 21 ¿ 4 G¿ ¿ ¿ + 83 ¿ 3 G¿ ¿ + 12 ¿ 2 G¿ + º 2 23 ¿ 3 G ¿ ¿ + 12 ¿ 2 G ¿ :

To understand the solution, let us use (A.15) to write Q2 in a convenient way,

(A.20) Q 2 = 12 º 21 ¿ 2 x2 G ¿ ¿ + 1 2 2
º (x
12 1
¡ ¿ )¿ G ¿ + 16 º 2 (2x 2 + ¿ )¿ G¿ :

Then through O(²2 ), our solution is

~ = Q 0 + ²Q1 + ² 2 Q2 + : : :
Q
(A.21) ¡ 4º 2 + º 21 2º 2 ¡ º 21 ¢
= G + ²º 1 ¿ xG¿ + 12 ²2 º 21 ¿ 2 x2 G ¿ ¿ + ² 2 x2 + ¿ ¿ G¿ ;
12 12
which we can re-write as

(A.22a) ~ ; x) = G(~¿ ; x);

Q(¿

with
£ ¡ 4º 2 + º 21 2 2º 2 ¡ º 21 ¢ ¤
(A.22b) ~¿ = ¿ 1 + ²º 1 x + ²2 x + ¿ +::: :
12 12
To obtain the option value, recall that

(A.23a) ~ ; x) = D(t; ts)²G(~

V (t; f ) = D(t; ts )²Q(¿ ¿ ; x)

and note that

(A.23b) ¿ ; x) = G(²2 ~¿ ; ²x) = G(A2 (K)~

²G(~ ¿ ; f ¡ K):

Thus the value of the option is

(A.24a) V (t; f ) = D(t; ts )G(¿ ¤ ; f ¡ K);

where
£ 4º 2 + º 21 2º 2 ¡ º 21 2 ¤
(A.24b) ¿ ¤ = A2 (K)¿ 1 + º 1 (f ¡ K) + (f ¡ K)2 + A (K)¿ + : : : :
12 12

A.2 Implied Volatility

Although (A.24) gives a closed form expression for the price of the option, it is not very convenient.
So we compute the equivalent Black volatility implied by this price. To simplify the calculation, we
take the square root of (A.24b) and obtain
p p h 2º 2 ¡ º 21 2º 2 ¡ º 21 2 i
(A.25) ¿ ¤ = A(K ) ¿ 1 + 12 º 1 (f ¡ K ) + (f ¡ K) 2 + A (K)¿ + : : : :
12 24
Recall
p that º 1 = A0 (K)=A(K), º 2 = A00 (K)=A(K), so the …rst two terms of (A.25) are
¿ [A(K) + 12 A0 (K)(f ¡ K) + : : : ]. This suggests expanding A around the average

(A.26) f av = 12 (f + K )

6
instead of K. Accordingly, we de…ne

(A.27) °1 = A0 (fav )=A(fav ); °2 = A00 (fav )=A(fav ); :: : ;

and re-write (A.25) in terms of °1 and ° 2 instead of º 1 and º 2 . This then shows that the option
price is

(A.28a) V (t; f ) = D(t; ts )G(¿ ¤ ; f ¡ K);

with
p p h ° ¡ 2° 21 2° ¡ ° 21 2 i
(A.28b) ¿ ¤ = A(fav ) ¿ 1 + 2 (f ¡ K )2 + 2 A (f av )¿ + : : : :
24 24
Now suppose we had started with Black’s model

(A.29) dF (t) = ¾ B F (t)dW

instead of dF = ®(t)A(F )dW . Repeating the preceding analysis for the special case ®(t) = ¾ B ; A(F ) = F ,
shows that the option price is

(A.30a) V (t; f ) = D(t; ts)G(¿ B ; f ¡ K);

with
p p h (f ¡ K) 2 ¾ 2B (tex ¡ t) i
(A.30b) ¿ B = ¾ B fav tex ¡ t 1 ¡ 2
¡ + : : : :
12fav 24
Since G(¿ B ; f ¡ K) is an increasing function of ¿ B , the Black price (A.30) matches the correct price
(A.28) if and only if
p p
(A.31) ¿ B = ¿ ¤:

Equating yields the implied volatility

(A.32a)
A(fav ) n 2 (f ¡ K)2 1 a2 A2 (f av )(tex ¡ t) o
¾B = a 1 + (°2 ¡ 2° 21 + 2 ) + (2° 2 ¡ °21 + 2 ) + : :: ;
fav fav 24 f av 24
where
A0 (fav ) A00 (f av )
(A.32b) fav = 12 (f + K); °1 = ; °2 = ;
A(f av ) A(fav )
and
Z tex
1
(A.32c) a= ®2 (t0 )dt0 :
tex ¡ t t

Setting t = 0 at f = F 0 yields (1.5).

Equation (A.32) gives the implied volatility for the European call option for the model

(A.33) dF (t) = ®(t)A(F )dW (t):

A similar analysis shows that the implied volatility for a European put option is given by the same
formula.

7
B Higher order results
The above analysis can be carried out to arbitrarily high order. Carrying it out through O(²4 ) yields
the more accurate (but more complicated) equivalent vol formula

(B.1)
aA(fav ) n ¢¡ ¢ µ2 ¡ ¢
¾B = 1+ 2° 2 ¡ °21 + f 12 + °2 ¡ 2° 21 + f22
f av 24 av 24 av

2 ¡ 2 ¢
¢ 10°2 ¡ 5° 21 + 5=f av
+ 2°4 + 4° 1 °3 + 3° 22 ¡ 3°21 ° 2 + 34 ° 41 ¡ 4f34 + 2
480 av 2f av
2 ¡ 2 ¢
¢µ 35°2 ¡ 40° 21 + 40=fav
+ 6° 4 ¡ 18° 1 ° 3 + 14° 22 ¡ 29° 21 °2 + 11°41 ¡ f114 + 2
2880 av fav
4
µ ¡3 2
5° 2 ¡ 10° 1 + 10=fav2 ¢ o
2 2 4 8
+ ° ¡ 6° ° ¡ 2° + 17° ° ¡ 8° + 4 + + : : : :
1440 4 4 1 3 2 1 2 1 fav f av
2

Here

(B.2a) f av = 12 (f + K ); µ = f ¡ K;

³ Z tex ´1=2
1
(B.2b) a= ®2 (t0 )dt0 ; ¢ = a2 A2 (f av )(tex ¡ t);
tex ¡ t t

and
A0 (fav ) A00 (f av ) A0 00(f av ) A0000 (fav )
(B.2c) °1 = ; °2 = ; °3 = ; °4 = :
A(fav ) A(fav ) A(fav ) A(f av )

For the special case of a power-law model, dF = ®(t)F ¯ dW , the equivalent Black vol reduces to

(B.3)
a n (1 ¡ ¯)2 ¢ µ2 (1 ¡ ¯)2 ¢2
¾B = 1¡¯
1+ 2
+ (1 ¡ ¯)(2 + ¯ ) 2
+ 4
(7 ¡ 54¯ + 27¯ 2 )
fav 24f av 24fav 1920fav
(1 ¡ ¯)2 ¢µ 2 (1 ¡ ¯)µ 4 o
+ 4
(29 ¡ 13¯ ¡ 16¯ 2 ) + 4
(72 + 34¯ ¡ 9¯ 2 ¡ 7¯ 3 ) + : : : :
2880f av 5760f av

References
[1] F. Black, “The Pricing of Commodity Contracts," J. Financial Economics 3 (1976), pp. 167-79

[2] J. Cox, “Notes on Option Pricing I: Constant Elasticity of Variance Di¤usion,” (Working paper,
Stamford University, Stanford Calif., 1975)

[3] B. Dupire, “Pricing with a Smile," RISK 7 (1994), pp. 18-20

[4] B. Dupire, “Pricing and Hedging with Smiles," Working Paper, Paribas Capital Markets (1993)

[5] P. S. Hagan and D. E. Woodward “Markov Interest Rate Models,” preprint (1997)

8
 [6] J. C. Hull, Options, Futures, and Other Derivatives (Prentice-Hall, Upper Saddle River, 1997)

[7] J. M. Harrison and D. Kreps, “Martingales and arbitrage in multiperiod securities markets,” J.
Econ. Theory 20 (1979), pp. 381-408

[8] J. M. Harrison and S. R. Pliska, “Martingales and stochastic integrals in the theory of continuous
trading,” Stoch. Proc. Applications 11 (1981), pp. 215-260

[9] J. Kevorkian and J. D. Cole, Perturbation Methods in Applied Mathematics (Springer-Verlag,

New York, 1981)

[10] P. Wilmott, J. Dewynne, and S. Howison, Option Pricing (Oxford Financial Press, Oxford,
1993)

9
Figure 1: Error in the equivalent vol for a normal model (¯ = 0). Shown is the di¤erence between
the equivalent vol given by (1.5) and the exact implied volatility as a function of the strike K for 1
year, 2 year, 5 year, and 10 year options. The case shown has F0 = 100 and an at-the-money Black
vol of 20%. The graph is scaled so that a 1 bp error would change 20% to 20.01%.

Figure 2: The error in the option price caused by using the equivalent vol formula (1.5). Shown is
the error as a function of the strike K for 1 year, 2 year, 5 year, and 10 year options. Here 1 bp
corresponds to an error of 10 ¡4 F0 for the same case as Figure 1.

10
Figure 3: The error in the option price caused by using the equivalent vol formula (1.5). Shown is
the ratio of the error to the time value of the option as a function of the strike K for 1 year, 2 year,
5 year, and 10 year options. Same case as Figure 1. Note that when the error is an appreciable
percent of the time value, the time value of the option is near zero.

Figure 4: Error in the equivalent vol for a normal model (¯ = 0). Shown is the di¤erence between
the more accurate equivalent vol given by (B.3) and the exact implied volatility as a function of the
strike K for 1 year, 2 year, 5 year, and 10 year options. Same case as Figure 1. Note that the graph
is scaled so that a 2% error would change 20% to 20.0002%.

11