1 The Black-Scholes Formula for a European Call or

Put

Recall:
V (f ) = e−r(T −t) ERN [f (ST )]
where the expectation is taken with respect to the risk-neutral measure.
In a risk-neutral world, the stock price dynamics is
√
ST = St e(r− 2 σ )(T −t)+σ
1 2
T −tZ
, Z ∼ N (0, 1)
or equivalently µ ¶ ·µ ¶ ¸
ST 1 2 2
log ∼N r − σ (T − t) , σ (T − t)
St 2
Note that f (ST ) is the payoff, a know function of ST , e.g.,

1.1 Evaluation of European Options

Evaluation of a European Call/Put at t = 0. Let us quote the results first:

c[S0 , T, K] = S0 N (d1 ) − Ke−rT N (d2 ) ,

p [S0 , T, K] = Ke−rT N (−d2 ) − S0 N (−d1 )
where
Z x
1 y2
N (x) = √ e− 2 dy
2π −∞
" #
1 S0 e(r+ 12 σ 2 )T
d1 = √ log
2
σ T K
" #
S0 e(r− 2 σ )T
1 2
1
d2 = √ log
σ2T K

1
Note that √
d2 = d1 − σ2T
First, let us evaluate the expectation of the following function
½ ax
e , x≥k
f (x) ≡
0, otherwise

where X is a Gaussian-distributed random variable with mean m and variance σ02 :

Z ∞
1 (x−m)2
E [f (x)] = f (x) √ e− 2σ02 dx
02
−∞
Z ∞2πσ
1 (x−m)2
= √ eax e− 2σ02 dx
2πσ 02 k
Complete the square:
2
(x − m)2 1 2 02 [x − (m + aσ 02 )]
ax − = am + aσ −
2σ 02 2 2σ 02
therefore,
Z ∞
2
[x−(m+aσ02 )]
am+ 12 a2 σ02 1 − 0
E [f (x)] = e √ e 2σ 2 dx
0
σ 2π k
Changing variable,
x − (m + aσ 02 )
y≡ ,
σ0
yields
Z∞
am+ 12 a2 σ 02 1 −y2
E [f (x)] = e √ e 2 dy
2π
k−(m+aσ 02 )
κ= σ
Z −κ
am+ 12 a2 σ 02 1 −y2
= e √ e 2 dy (even-symmetry of a Gaussian)
−∞ 2π
µ ¶
am+ 12 a2 σ 02 k − (m + aσ 02 )
= e N −
σ0

Therefore, we have
1 2 02 −k + m + aσ 02
E [f (x)] = eam+ 2 a σ
N (d) , d≡ (1)
σ0

2
1.1.1 European Call
Applying Eq. (1) to a European call:
Z ∞ (x−(r− 12 σ2 )T )
2

rT x 1 −
V (f ) = e (S0 e − K)+ √ e 2σ 2 T dx
−∞ 2πσ 2 T
Note that
K
S0 ex − K ≥ 0 =⇒ x > log
S0
1. For the first term in the payoff, i.e., S0 ex , we use the result above with
µ ¶
K 1 2
a = 1, k = log , m = r − σ T, σ 02 = σ 2 T
S0 2
therefore,
Z ∞ (x−(r− 12 σ2 )T )
2

−rT x1 −
e S0 e √ e 2σ 2 T dx
k 2πσ 2 T
= S0 e−rT e(r− 2 σ )T + 2 σ T N (d1 )
1 2 1 2

= S0 N (d1 )
where
¡ ¢
− log SK0 + r − 12 σ 2 T + σ 2 T
d1 = √
2
" σ T1 2 #
1 S0 e(r+ 2 σ )T
= √ log
σ2T K

2. For the 2nd term (i.e., -K), choose

a = 0,
then,
Z ∞ (x−(r− 12 σ2 )T )
2
1
e−rT
K√ e− 2σ 2 T dx = Ke−rT N (d2 )
2
2πσ T
k
where
¡ ¢
− log SK0 + r − 12 σ 2 T
d2 = √
σ2T
" #
S0 e(r− 2 σ )T
1 2
1
= √ log
σ2T K

Therefore,
c (S0 , T, K) = S0 N (d1 ) − Ke−rT N (d2 )

3
1.1.2 European Put

How to evaluate a put? Use the put-call parity

p − c = Ke−rT − S0

Therefore,

p (S0 , T, K) = c (S0 , T, K) + Ke−rT − S0

= S0 N (d1 ) − S0 + Ke−rT − Ke−rT N (d2 )
= −S0 (1 − N (d1 )) + Ke−rT (1 − N (d2 ))
= −S0 N (−d1 ) + Ke−rT N (−d2 )

Hence
p (S0 , T, K) = Ke−rT N (−d2 ) − S0 N (−d1 )
Note that the notation T can be understood as the maturity of the contract counting
from the day when the option is setup or it can also be understood as the time-to-maturity
– which is, sometimes, emphasized through the notation T − t with T being reserved for
maturity.

Note that

1. These prices are good as long as the lognormal stock price dynamics is a good model
for our market;
2. Parameters in the formula:

S0 – the present value of a stock,

K – Strike,
r – risk-free interest rate,
T – maturity or time-to-maturity
σ – volatility

what is the value of σ?

4
 (a) Historical volatility;
(b) Implied volatility – cf. Volatility smile, volatility skew.

Issues:

1. How good is the lognormal dynamics?

2. How to hedge away some of the problem?

2 Hedging

2.1 Hedging in a Binomial World

· ¸
∂f 1 2 2 ∂2f
f0 = f (S0 , t + δt) + −rf (S0 , t + δt) + rS0 + σ S0 2 δt + o (δt)
∂S0 2 ∂S0
where all the derivatives are evaluated at t + δt.

Suppose we misparameterized σ :

σ 0 = σ + δσ

where δσ is the error.

5
 Then, the incorrect price for our contingent claim:
· ¸
0 ∂f ∂f 1 2 2 ∂2f
f0 = f (S0 , t + δt) + δσ + −rf (S0 , t + δt) + rS0 + σ S0 2 δt + o (δt)
∂σ ∂S0 2 ∂S0
For the purpose of illustrating the idea of hedging, we have assumed δσ and δt are of the
same order, otherwise, there are further expansions of those derivatives with respect to δσ.
Here, we neglect higher order terms, e.g. O (δtδσ) .
Therefore, the mispriced amount is
δf0 ≡ f00 − f0
∂f
≈ δσ
∂σ
However, if we have another contingent claim on the same stock to form a portfolio:
f + xg, x – the number of unit of g-option.
Then the total mispricing will be
µ ¶
∂f ∂g
δ (f + xg) ≈ +x δσ
∂σ ∂σ
if
∂f ∂g
+x = 0
∂σ ∂σ
∂f
−
i.e., x = ∂σ
∂g
∂σ
then we can hedge away potential mispricing due to incorrect volatility parameterization to
O (δσ) – a Vega hedging.

Terminology:
V ega :
∂f ∂g
Vf ≡ , Vg ≡
∂σ ∂σ
i.e.,
Vf
x=−
Vg

Question: Have we hedged away all risks yet? Let’s analyze this issue further.

Recall that the replicating portfolio in a risk-neutral valuation is

−f0 + ∆f S0 + Bf = 0
| {z }
replicating prtflo

6
in a correctly parameterized world. i.e.,
−f+ + ∆f S+ + Bf erδt = −f− + ∆f S− + Bf erδt
f+ − f−
∴ ∆f =
S+ − S−
is the amount of stock needed to hedge away the risk
Due to σ-misparameterization:
−f00 + ∆0f S0 + Bf0 = 0 at t = 0
therefore,
f+0 − f−0
∆0f =
S+ − S−
0 0 0 0
where f+ , f− and Bf are computed using σ . So in a time-step δt, our risk is
¡ ¢ ¡ ¢
δΠf = −f+ + ∆0f S+ + Bf0 erδt − −f− + ∆0f S− + Bf0 erδt
⇑
N.B. in the real world, our f has to pay f+
rather than f+0
∴ δΠf = − (f+ − f− ) + ∆0f (S+ − S− )

∵ ∆f = ∆f (σ)
∂∆f
∴ ∆0f ≈ ∆f + δσ + o (δσ)
∂σ
therefore,
∂∆f
δΠf = − (f+ − f− ) + ∆f (S+ − S− ) + δσ (S+ − S− )
| {z } ∂σ
=0
∵ perfectly hedged
with correct σ
∂∆f
∴ δΠf = (S+ − S− ) δσ
∂σ
which, in general, is not zero. However,
³ √ ´
S+ − S− ≈ O σ δt
µµ ¶ ¶
∂∆f √
∴ δΠf = O σ δt δσ
∂σ
which contains risks – Either we are content to live with these risks (they could be small
∂∆
or large, depending on the combination of ∂σf σδσ) or we can try to hedge further – Let’s
see how theoretically this can be done. First note that, even for our portfolio
f + xg

7
we have µ ¶
∂∆f ∂∆g
δΠf +xg = δΠf + δΠg = +x (S+ − S− ) δσ
∂σ ∂σ
which means there are still risks, i.e. our portfolio f + xg is not completely ∆-neutral.
Now suppose we have another contingent claim on the same stock to form a new portfolio:
f + xg + yh
we can choose x, y, such that
∂f ∂g ∂h
V ega-Neutral: +x +y =0 i.e. Vf + xVg + yVh = 0 (2a)
∂σ ∂σ ∂σ
∂∆f ∂∆g ∂∆h
and ∆-Neutral: +x +y =0 (2b)
∂σ ∂σ ∂σ
then we have hedged away potential mispricing and risks due to misparameterization of σ.

Importance of being nonlinear: Question: can we use the stock as our third option for
hedging, i.e.,
h (ST ) = ST
i.e., the stock itself for our option h (S)? Note that

∂S
Vh = = 0 – a stock has vanishing V ega
∂σ · ¸
∂S0 S+ − S−
and ∆h = =1 or ∆h = =1
∂S S+ − S−
i.e., a stock has ∆ = 1
=⇒ stock S is a linear derivative
Can we use the stock itself for our h?
Since Eqs. (2) now become
∂f ∂g
+x +y·0 = 0
∂σ ∂σ
∂∆f ∂∆g
and +x +y·0 = 0
∂σ ∂σ
leading to no solution for x, and y, in general.
The story is just to give you some sense of how issues of hedging arise and how hedging
can be done. This simple example illustrates the need for nonlinear derivatives for hedging
purposes.
Conclusion:
Even if a stock price dynamics is not 100% accurate, as long as it is sufficiently close to
the true dynamics – meaning both model specification and model parameterization – then
we can use a well-balanced (hedged) portfolio to eliminate most of risks.

8
2.2 Hedging (General Formulation) – Greeks
2.2.1 Greeks

Portfolio value:
Π = Π (t, S, σ, r)
where t is time-to-maturity. Then
∂Π ∂Π ∂Π ∂Π
δΠ = δt + δS + δσ + δr
∂t ∂S ∂σ ∂r
1 ∂2Π
+ (δS)2 + · · ·
2 ∂S 2
where δt indicates the changing of time. What are the Greeks? They are no more than
∂Π
T heta : Θ= ,
∂t
∂Π
Delta : ∆= ,
∂S
∂Π
V ega : V= ,
∂σ
∂Π
rho : ρ =
∂r
∂2Π
Gamma : Γ=
∂S 2
Note that a portfolio contains, e.g., stocks, calls, puts, etc. each of which has its own
corresponding ∆, Γ,etc. For example, for a stock,
∆S = 1
ΓS = 0
VS = 0

2.2.2 Greeks for a European Call/Put:

∂
∆c = c (S0 , T, K) = N (d1 ) (How to evaluate? HW)
∂S0
∂
∆p = p (S0 , T, K) = −N (−d1 ) = N (d1 ) − 1
∂S0
the second line of which can be seen directly from the put-call parity.
Note that the hedging portfolio in a risk-neutral way would be
∆c S0 − c
With changing ∆c , one has to rebalance.

9
 ∂
∆ := – sensitivity to the change of the stock price.
∂S0
Call: 1 ≥ N (d1 ) ≥ 0 Put: − 1 ≤ N (d1 ) − 1 ≤ 0

Q: Why is it so difficult to hedge a cash-or-nothing?

∂2
Γ := – ∆-sensitivity to the change of S0
∂S02
· 2¸
1 d
Call: Γc = √ exp − 1 > 0 Put: Γp = Γc
S0 2πσ 2 T 2

Γp = Γc can be seen directly from the put-call parity.

10
 ∂
V := – sensitivity to the volatility change
∂σ
r · 2¸
T d
Call: Vc =S0 exp − 1 > 0, Put: Vp = Vc
2π 2

again, the relation Vp = Vc can be seen from the put-call parity.

Note that the Black-Scholes formula assumes a constant σ. But, if volatility changes,
then rebalance is needed.

11
 ∂
Θ := – sensitivity to "time-to-maturity"
∂t
· 2¸
S0 σ d
Call : Θc = − √ exp − 1 − rKe−rT N (d2 ) < 0
2 2πT 2
· 2¸
S0 σ d >0
Put : Θp = − √ exp − 1 + rKe−rT N (d2 ) = {
2 2πT 2 <0

What do we mean by sensitivity to "time-to-maturity" since there is nothing uncertain

about time after all? Actually this is another way of looking at Gamma Γ.

Recall the Black-Scholes PDE:

∂f 1 2 2 ∂ 2 f ∂f
+ σ S 2
+ rS − rf = 0
∂t 2 ∂S ∂S
i.e.,
1
Θ + σ 2 S 2 Γ + rS∆ = rf (3)
2
Note that

1. If we need nothing to enter a contract (a portfolio with many options), then

f (0) = 0

If we want to maintain that way, i.e., constant rebalance to ensure f = 0, then, Eq.
(3) yields
∆-neutral, Θ-neutral =⇒ Γ-neutral

12
 2. For a ∆-neutral portfolio, we have
1
Θ + σ 2 S 2 Γ = rf
2
then
large, positive Θ =⇒ large, negative Γ
large, negative Θ =⇒ large, positive Γ

∂
ρ := – sensitivity to the change of interest rate
∂r
Call : ρc = T Ke−rT N (d2 ) > 0
Put : ρp = −T Ke−rT N (−d2 ) = T Ke−rT [N (d2 ) − 1] < 0

2.2.3 General Hedging:

∆-Hedge: e.g.,
n1 units of an option f1 on the same stock.
and the portfolio is
Π1 = n1 f1 + ns S0 + B
where B stands for the amount of bond. ∆-neutral is
∂
Π1 = 0
∂S0

i.e., n1 ∆1 + ns = 0
which is insensitive to the change of the stock price. Since,
n1 f1 + ns S0 + B = V1
n1 ∆1 + ns = 0
which are two equations with two unknowns, n1 , ns , there is no more freedom to hedge other
Greeks.

13
V-Hedge: However, if we have two options f1 , f2 (on the same stock) with n1 and n2 units,

respectively, then the value of the portfolio is

Π2 = n1 f1 + n2 f2 + ns S0 + B
its value is
n1 f1 + n2 f2 + ns S0 + B = V2 , (4)
with ∆-neutral position, i.e.,
∂Π2
= 0
∂S0
=⇒ n1 ∆1 + n2 ∆2 + ns = 0 (5)

Now we have more freedom for hedging. For example, we can demand V ega-neutral, i.e.,
∂
Π2 = 0 =⇒
∂σ
n1 V1 + n2 V2 = 0 (6)

Eqs. (4) , (5) and (6) can be solved for the three unknowns, n1 , n2 , n3 – thus, achieving
∆-neutral and V ega-neutral, reducing the exposure to the changes or mis-specification of
volatility, etc.

Note that

1. By increasing types of options, we can hedge away other kinds of risks described by
other Greeks.

2. Dynamic balancing: Often times Γ and V ega are monitored but not zeroed out. ∆ is
zeroed out daily by rebalancing shares.

3. There is a difficulty of Γ-neutral and V-neutral – which require nonlinear derivatives

that are traded at competitive prices.

2.2.4 Speculation using Greeks

Consider only a European call

∂c
δc = δS = ∆δS
∂S
δc ∆ S∆ δS
= δS =
c c c S

14
If
S
∆À1
c
then, a percentage change of stock price δS/S can lead to an appreciably large percentage
change of option price.

One can also bet on volatility:

δΠ
δΠ = δσ = Vδσ
δσ µ ¶
δΠ V δσ
= σ
Π Π σ

If a portfolio is constructed in such a way that

V
σÀ1
Π
then, it is possible to use this portfolio for speculation with high leverage, i.e., a change of
V
volatility is magnified by a factor Π σ in the portfolio price.

15