Stock, Implied, Local Volatilities and Black Scholes Pricing.

Ilya I. Gikhman
6077 Ivy Woods Court Mason,
OH 45040, USA
Ph. 513-573-9348
Email: ilya_gikhman@[Link]

JEL : G12, G13

Key words. Black Scholes pricing, implied volatility, local volatility.

Abstract. In this paper we present a critical point on connections between stock volatility, implied
volatility, and local volatility. The essence of the Black Sholes pricing model is based on assumption that
option piece is formed by no arbitrage portfolio. Such assumption effects the change of the real
underlying stock by its risk neutral counterpart. Market practice shows even more. The volatility of the
underlying should be also changed. Such practice calls for implied volatility. Underlying with implied
volatility is specific for each option. The local volatility development presents the value of implied
volatility.

Implied volatility (IV) considers inverse construction. It was heuristically ‘by definition’ assumed that
option data correspond to BS model with underlying stock having undefined volatility with risk free drift
coefficient. Estimates of volatility show that the volatility of the heuristic underlying does not have any
relationship to the stock volatility. In [5] they presented calculations that highlighted the fact that each
option has a specific implied volatility. For example call and put options at the same date, maturity, and
strike price have different implied volatilities. Formally it means that market prices of the options do not
follow BS pricing model. On the other way they decided to define implied volatility process. It is
common to defined implied volatility with the help of inverse function of the option price with respect to
volatility sigma. Such way should return underlying sigma. If we use stock then invers function will
return stock sigma. If we use implied volatility process then its sigma is undefined and inverse function
definition does not have sense. Instead we use option data to present implied volatility by using Local
Volatility (LV) concept.

Electroniccopy
Electronic copyavailable
available at:
at: [Link]
[Link]
 I. In this section we briefly recall Black Scholes option pricing approach. In the theory we deal

with a security which follows a Geometric Brownian motion equation

dS ( t ) = μ ( t ) S ( t ) dt +  ( t ) S ( t ) dw ( t ) (1)

Here μ ( t ),  ( t ) are known deterministic drift and volatility coefficients and w ( t ) is a Wiener process
defined on original probability space {  , F , P }. No arbitrage option price at a moment t , t ≥ 0 is
defined based on a construction of the instantaneous risk free portfolio П ( t ) over the infinitesimally
small interval [ t , t + dt ). Assume that price of the call option C ( t , S ) is a sufficiently smooth
deterministic function in t , S. The option price C ( t , S ) depends also on option contract parameters such
as a maturity date T , T ≥ t and strike price K. Call option contract is formally defined by its payoff at
maturity T

max { S ( T ) – K , 0 }

which represents an option but not obligation to buy European call option for K. Option price is defined
synthetically. Let us briefly recall the construction of the option price [4]. Define a perfectly hedged
portfolio at a moment t. It is formed by long call option and a Δ ( t ) portion of underlying stocks

Π ( u , t ) = C ( u , S ( u )) – Δ ( t ) S ( u ) (2)

Here t is a fixed moment of time and variable u ≥ t. Let

/
Δ(t)= C S ( t , S ( t ))

Applying Ito formula it is easy to verify that change in the value of the portfolio at t does not hold a
market risk associated with the ‘dw ( t )’ term , i.e.

 C ( t ,S( t ) ) S 2 ( t ) σ 2  2 C ( t ,S( t ) )
d Π ( u , t ) | u=t = d Π ( u , S ( u ) ) | u=t = [  ] dt
t 2 S 2

Bearing in mind that the right hand side of the latter formula is nonrandom at t we conclude that

d Π ( u ,t ) | u = t = r Π ( u , t ) dt | u = t

From latter equality it follows that call option price is a solution of the Cauchy problem

/ 1 //
C t ( t , x ) + r x C /x ( t , x ) + C xx ( t,x)σ2x2 – rC(t,x) = 0 (BSE)
2

with boundary condition

C ( T , x ) = max { x – K , 0 }

The solution of the (BSE) equation can be represented in the form

C ( t , x ) = E exp – r ( T – t ) max { S r ( T ; t , x ) – K , 0 } (3)

Electroniccopy
Electronic copyavailable
available at:
at: [Link]
[Link]
where S r ( u ; t , x ) , u > t is the solution of the stochastic Ito equation

d Sr ( u ) = r Sr ( u ) d u + σ Sr ( u ) d w ( u ) (4)

and S r ( t ) = x . Stochastic process (4), called also risk neutral process is usually interpreted as
underlying of the Black Scholes option pricing. Risk neutral random process S r ( t ) does not a represent
traded security. It can be seen as auxiliary random process.

In general definition of the derivatives price a derivative instrument is defined by the values of its
underline security. Such vision is based on a long market trading experience. On the other hand Black
Scholes pricing construction leads to the random process (4). There exists the risk neutral concept which
attempts to cover or also to hide existing contradiction.

Pricing formula (3) shows that actual underlying of the call option is the random process S r ( u ) on
original probability space{ Ω , F , P }. Risk neutral concept suggests to consider asset price defined by
equation (1) on the risk-neutral probability space { Ω , F , Q } where the risk-neutral probability measure
Q is defined on -algebra F by the formula

T T
μ  r 1 μ  r 2
Q(A) = 
A
{ exp 
0
[
σ
dwQ(t) –
2 
0
(
σ
) dt ]}P(dω)

where A  F and w Q ( t ) denotes a Wiener process on { Ω , F , Q }. Then the random process S r ( t ) is a

solution of the risk-neutral equation (4) on probability space { Ω , F , P }which has the same distribution
as original stock S ( t ) on risk neutral probability space { Ω , F , Q } with Wiener process

t
μ  r
w Q( t ) = w ( t ) + 
0
σ
du

Thus the essence of the risk neutral valuations is to put the real stock equation on a probability space
{ Ω , F , Q }. Such construction is applied for representation of the dependence of the call option price on
real stock. Nevertheless the solution of the BSE does actually depend on S r ( t ) but not on S ( t ). Recall
that regardless of whether options are exist or not the stock equation (1) is defined on original probability
space { Ω , F , P }. One can avoid risk neutral interpretation dealing with S r ( t ) on original probability
space { Ω , F , P }. The solution of the equation (4) can be written as

t t
1 2
S r ( t ; 0 , x ) = x exp 
0
(r –
2
 )dl + 
0
 dw(l)

and Black Scholes equation solution (3) can be represented in the following form

C ( t , x ) = x N ( d 1 ) – K exp – r ( T – t ) N ( d 2 ) (5)

where

Electroniccopy
Electronic copyavailable
available at:
at: [Link]
[Link]
 1 x σ 2 (T - t )
d1 = [ ln + ] , d2 = d1 –  T-t
σ Tt K exp - r ( T - t ) 2

Here N (  ) denotes the cumulative distribution function of the standard Gaussian random variable with
parameters ( 0 , 1 ). Consider time evolution of the call option. For each market scenario  the value of
the option at a date t  [ 0 , T ) is C ( t , S ( t )) = C ( t , S ( t ) ; T , K ). Applying Ito formula and bearing
in mind (1) we arrive at

t
/
C ( t , S ( t )) = C ( 0 , x ) +  [C t ( u , S ( u )) + C /x ( u , S ( t )) μ ( u ) S ( u ) +
0

t
1 //
+ C xx (
2 2
u , S ( u )) σ ( u ) S ( u ) ] du +  C /x ( u , S ( u ))  ( u ) S ( u ) dw ( u )
2 0

For calculations we use a data C j = C ( t j , S j ; T , K ) , j = 0, 1, … n where 0 = t 0 < t 1 < … < t n = t that

is specified by a fixed scenario . In case when volatility  in (1) is a constant its estimate <  > is based
on observed data C j , j = 0, 1, … n which in general can show dependence of the estimate <  > on K
and T – t.

II. The call option pricing formula (5) depends on real stock volatility  =  ( t ). Recall that

only stock and option prices are observable variables. The BS model introduces option price as a function
of the variables ( t , S ) while ( T , K ; σ ) are considered as fixed parameters.

Making a heuristic assumption that market data are perfectly follow to the BSE solution and bearing in
mind that the BS price is an increasing function of volatility one can present volatility as a function of
historical prices. Such volatility is called implied volatility. It is implied by assumption that BS model
runs option market.

Consider now call option price as a function of the variable σ when t , S ; T , K are assumed to be fixed
parameters. One can verify that

d 2
S( t ) T  t exp 
 C ( t , S; T , K ; σ ) 2
= > 0 (6)
σ 2π

Therefore for each fixed values ( t , S ; T , K ) the function c = C ( t , S ; T , K ; σ ) has an inverse

function in σ

σ ( c ) = C–1 ( t , S ; T , K ; c ) (IV)

Here c is an admissible by the BS model value of the call option price. The volatility which corresponds
to the options observed prices given an assumption that market prices of the options are formed by the BS
formula is called implied volatility, (IV). Thus IV is served as an adjustment to BS theory to present the
correspondence of the BS option pricing theory and the real options market prices. One should remark

Electronic copy available at: [Link]

that in the theory formula (IV) should present volatility of the underlying stock. Such remark does not
work for IV concept which is trying to explain why the estimates of the volatilities of different options do
not relate to each other while in theory these volatilities are equal to the same stock volatility.

In order to develop new IV approach we should ignore the real stock equation (1) which approximate the
real stock prices and focus on implied volatility equation

d Siv ( t )
= r ( t ) d t + σ iv ( t , S iv ( t )) d w ( t ) (7)
Siv ( t )

2
It will be clear that σ iv ( t , x )   2 ( t ). One can note that σ iv ( t , S iv ) on right hand side (7) is not
formally defined. We do not have a definition of the function σ iv ( t , S ). It is unknown function.

Remark. Given S iv ( t ) formula (3) should be rewritten as

C ( t , x ; T, K ) = E exp – r ( T – t ) max { S iv ( T ; t , x ) – K , 0 } (8)

IV in the formula (8) should correspond either BS theory as well as trading experience. From (8) it
follows that the function C ( t , x ; T, K ) should be represented by a IV-version of the BSE, i.e.

/ 1 //
C t ( t , x ) + r x C /x ( t , x ) + C xx ( t,x)σ 2
iv (t,x)x2 – rC(t,x) = 0 (IVBSE)
2

Now we arrive at two versions of the BSEs. One (3) corresponds to the real stock which specified by
equation (1) and other one which specified by the (7) implied volatility function. Taking the difference of
two equations (BSE) and (IVBSE) and bearing in mind that x > 0 it follows that

// 2
C xx ( t,x) [σ iv (t,x) – 2(t)] = 0

From which it follows that C ( t , x )  0. Bearing in mind that the value C ( T , x ) = max { x – K , 0 }
does not equal to zero we arrived at contradiction that (BSE) and (IVBSE) could present the same
function C ( t , x ; T , K ). Therefore we should consider IV process (7) as the underlying of the option
and assume that real stock equation is relevant to BS pricing model

Actually the difference between real and implied stock prices justifies that market does not follow BS
pricing model and therefore the perfect hedging pricing approach does not reliable for trading options
practice.

III. In order to present local volatility approach we assume that there is no real world stock

equation (1). LV theory was originated by observations that implied Black-Scholes volatilities strongly
depend on the maturity and the strike price [1]. For the stock process (7) consider a set of the call options
with different maturities and strikes t ≤ T 1 < T 2 < … ; K 1 < K 2 < … and denote
c j i = C ( t , S ; T j , K i ; σ ). Then for each fixed pair of indexes j , i there exists inverse function for
which

σ ( t , S ; T j , K i ) = C – 1 ( t , S ; T j , K i ; c j i ) , j = 1, 2, … (9)

Electronic copy available at: [Link]

In [5] they presented the fact that call and put options with the same values ( t , S , T , K ) have different
implied volatility. This observation highlights the fact that implied volatility functions for calls, puts for
the same ( t , S ( t ), T , K ) are different. Hence trading practice it makes sense to reject the BS model.
But loyal to BS pricing developers chose other way. The LV concept assumes that BS pricing is perfectly
good represents market prices of the options and real stock should be exchanged for risk neutral implied
volatility process with unknown volatility coefficient. Note that as far as there is no such stock process in
the market the hedged portfolio which is underlying of the BS pricing does not make any real sense and
therefore ‘no-arbitrage’ pricing looks like financial theoretical fantasy.

Let us briefly recall details of the Local Volatility construction. In original Dupire papers the basic
process is given in the form (7). The LV or volatility smile was developed in [1-3]. Let us outline
mathematical framework of the LV model. Suppose that underlying of the call option is given by (7) and
S ( t ) = x. Bearing in mind equation (7) option price (3) should be rewritten as

C ( t , x ; T , K ) = E exp – r ( T – t ) max { S iv ( T ; t , x ) – K , 0 } (10)

Fix the variables ( t , x ) and consider call option price as a function of the variables ( T , K )  [ t , + ∞ )
× ( 0 , + ∞ ). Then



exp r ( T – t ) C ( t , x ; T , K ) = 
K
(y - K) p(t,x;T,y)dy (11)

where p ( t , x ; T , y ) is probabilistic density of the stochastic process S iv ( T ; t , x ). For simplicity

notations we will omit next the low index ‘iv’ implied volatility function S and let us assume that density
is a sufficiently smooth function of its variable. Twice differentiation in (11) with respect to K leads us to
a well-known equality

//
exp r ( T – t ) C KK (t,x;T,K) = p(t,x;T,K) (12)

Density function p ( t , x ; T , y ) of the diffusion process (7) satisfies first Kolmogorov equation also
known as the Fokker-Plank equation

  1 2 2
p(t,x;T,K) = – [r p(t,x;T,K)] + [σ iv (T,K)K2p(t,x;T,K)]
T K 2 K 2

Note that

 2 2
C ( t , x ; T , K ) = exp r ( T – t ) [rC(t,x;T,K)] +
T K 2 K 2
2 
+ exp r ( T – t ) C(t,x;T,K),
K 2 T

 2 2 
[ exp r ( T – t ) r C ( t , x ; T , K ) ] = exp r ( T – t ) rC(t,x;T,K) ,
K K 2 K 2 K

Electronic copy available at: [Link]

 2 2 2
[ exp r ( T – t ) σ iv (T,K)K2 C(t,x;T,K)] =
K 2 K 2

2 2 2
= exp r ( T – t ) [σ iv (T,K)K2 C(t,x;T,K)].
K 2 K 2

Divided both sides of the equation by exponential factor and integrating twice in K we arrive at the
Dupire equation

  1 2 2 2
C(t,x;T,K) + r C(t,x;T,K) – σ iv (T,K)K C(t,x;T,K) +
T K 2 K 2

(13)

+ rC(t,x;T,K) = 0

In this equation option price is a known function and implied volatility is unknown. Therefore

 C ( t , x ; T, K )  C ( t , x ; T, K )
 r[  C ( t , x ; T, K ) ] 1
T K 2
σ iv ( T , K ) = { } (14)
1 2  2 C ( t , x ; T, K )
K
2 K 2

Here variable T ≥ t and K > 0 and we took into account discount factor. Formula (13) represents diffusion
coefficient in equation (7) which specifies the BS option price with implied volatility. Formula (14) for
is a final conclusion if the IV-LV development.

We can also make an extension of the IV-LV construction. Using formula (12) we can present formulas
for implied volatility in (7). The system of Kolmogorov forward and backward equations can be used to
define volatility of the implied stock equation (7). Similar to (13) we arrive at equation

 
C ( t , x ; T , K ) + μ iv ( T , K ) C(t,x;T,K) – (15)
T K

1 2 2
– σ iv (T,K)K2 C(t,x;T,K) + rC(t,x;T,K) = 0
2 K 2

On the other hand we can also apply backward Kolmogorov equation. Then

  1 2 2
2
p(t,x;T,K) = – [r(t)p(t,x;T,K)] – σ iv (t,x)x p(t,x;T,K)
t x 2 x 2

Using formula (12) we can change the order of derivatives with respect to x and K. Then integrating twice
in K and cancelling exponent factor lead us to equation

Electronic copy available at: [Link]

  
C(t,x;T,K) + r(t) C(t,x;T,K) + (16)
t x

1 2 2
+ σ iv (t,x)x2 C(t,x;T,K) – r C(t,x;T,K) = 0
2 x 2

Note that term r C corresponds to discount factor in BS formula (16). System (15), (16) defines unique
solution of implied volatility of the equation (7).

Conclusion. Implied volatility that is specified by the local volatility does not improve the BS pricing
model. It applies the not-real market underlying process of the type (7) for the BS model. The hedged
portfolio used as the basis of the non-arbitrage pricing does not have any sense as there is no asset in the
market that governed by the implied – local volatility function (7). Therefore the BS pricing does not have
a possibility of its practical applications. It can be used for theoretical or educational needs.

Besides, there is no answer of how the implied - local volatility equation (7) can explain the market
phenomena that call and put options with different K have different implied volatilities while they have
the same underlying that does not depend on K.

We should also remark that in the first papers [1-3] equation (7) did not presented in explicit form and
real stock equation (1) did not rejected from consideration. In such setting the stock process and the
heuristic process with local/implied volatilities represent an

ambiguity.

Electronic copy available at: [Link]

References.

1. Dupire, B., Pricing and Hedging with Smiles, Research paper, 1993, p.9.
2. Derman, E., and Kani,I., Riding on a smile, Risk, 7 (1994), p.32-33.
3. Dupire, B., Pricing with a Smile, Risk Magazine, 7, 1994, p.8-20.
4. Gikhman I., [Link]
5. Jarrow, R., Turnbull A., Derivative Securities, 2nd ed. South-Western College Publishing.

Electronic copy available at: [Link]