IEOR E4602: Quantitative Risk Management Spring 2016

c 2016 by Martin Haugh

A Brief Review of Derivatives Pricing & Hedging

In these notes we briefly describe the martingale approach to the pricing of derivatives securities. While most
readers are probably more familiar with the dynamic replication approach for pricing derivatives this approach
only works in complete market settings. It is therefore less general than the martingale approach. Moreover, the
approach taken in practice for pricing derivatives has more in common with the martingale approach. After
briefly describing the basic ideas of the martingale approach we will apply it to the pricing of call options in the
binomial model and discuss how the Black-Scholes price is obtained in the limit as the number of periods goes
to infinity.
We will also discuss the weaknesses of the Black-Scholes model, i.e. the geometric Brownian motion model, and
this leads us naturally to the concept of the volatility surface which we will describe in some detail. We will also
derive and study the Black-Scholes Greeks and discuss how they are used in practice to hedge option portfolios.

1 Martingale Pricing Theory for Multi-Period Models

We assume there are N + 1 securities, m possible states of nature and that the true probability measure is
denoted by P = (p1 , . . . , pm ) with each pi > 0. We assume that the investment horizon is [0, T ] and that there
are a total of T trading periods. Securities may therefore be purchased or sold at any date t for
(i)
t = 0, 1, . . . , T − 1. We use St to denote the time t price of the ith security for i = 0, . . . , N . Figure 1 below
shows a typical multi-period model with T = 2 and m = 9 possible states. The manner in which information is
revealed as time elapses is clear from this model. For example, at node I14,5 the available information tells us
that the true state of the world is either ω4 or ω5 . In particular, no other state is possible at I14,5 .

c ω1




c

Figure 1  ω2
 
c
 

hhhhhhh
 1,2,3 hhc ω3
 I1
c ω4


  
 

c c c

  ω5
Q
4,5
I0 Q
Q I1
c ω6
Q
Q
Q  c
Q (

(((( ω7
ch
Q  (
(
Q( hh(
H
6,7,8,9HHhhhhhhc
I1 H ω8
HH
H
HHd
ω9

t=0 t=1 t=2

A Brief Review of Derivatives Pricing & Hedging 2

Note that the multi-period model is composed of a series of single-period models. At date t = 0 in Figure 1, for
example, there is a single one-period model corresponding to node I0 . Similarly at date t = 1 there are three
possible one-period models corresponding to nodes I11,2,3 , I14,5 and I16,7,8,9 , respectively. The particular
one-period model that prevails at t = 1 will depend on the true state of nature. Given a probability measure,
P = (p1 , . . . , pm ), we can easily compute the conditional probabilities of each state. In Figure 1, for example,
P(ω1 |I11,2,3 ) = p1 /(p1 + p2 + p3 ). These conditional probabilities can be interpreted as probabilities in the
corresponding single-period models. For example, p1 = P(I11,2,3 |I0 ) P(ω1 |I11,2,3 ).

Trading Strategies and Self-Financing Trading Strategies

We first need to introduce the concepts of a trading strategy and a self-financing trading strategy. We will
assume initially that none of the securities pay “dividends” or have intermediate cash-flows.

Definition 1 A predictable stochastic process is a process whose time t value, Xt say, is known at time
t − 1 given all the information that is available at time t − 1.

(0) (N )
Definition 2 A trading strategy is a vector, θt = (θt (ω), . . . , θt (ω)), of predictable stochastic processes
that describes the number of units of each security held just before trading at time t, as a function of t and ω.
(i)
For example, θt (ω) is the number of units of the ith security held1 between times t − 1 and t in state ω. We
(i)
will sometimes write θt , omitting the explicit dependence on ω. Note that θt is known at date t − 1 as we
insisted in Definition 2 that θt be predictable. In our financial context, ‘predictable’ means that θt cannot
depend on information that is not yet available at time t − 1.

Example 1 (Constraints Imposed by Predictability of Trading Strategies)

Referring to Figure 1, it must be the case that for all i = 0, . . . , N ,
(i) (i) (i)
θ2 (ω1 ) = θ2 (ω2 ) = θ2 (ω3 )
(i) (i)
θ2 (ω4 ) = θ2 (ω5 )
(i) (i) (i) (i)
θ2 (ω6 ) = θ2 (ω7 ) = θ2 (ω8 ) = θ2 (ω9 ).

(i)
Exercise 1 What can you say about the relationship between the θ1 (ωj )’s for j = 1, . . . , m?

Definition 3 The value process, Vt (θ), associated with a trading strategy, θt , is defined by
 P
N (i) (i)
i=0 θ1 S0 for t = 0


Vt =
 PN θ(i) S (i) for t ≥ 1.

i=0 t t

Definition 4 A self-financing (s.f.) trading strategy is a strategy, θt , where changes in Vt are due entirely to
trading gains or losses, rather than the addition or withdrawal of cash funds. In particular, a self-financing
strategy satisfies
N
(i) (i)
X
Vt = θt+1 St for t = 1, . . . , T − 1.
i=0

Definition 4 states that the value of an s.f. portfolio just before trading or re-balancing is equal to the value of
the portfolio just after trading, i.e., no additional funds have been deposited or withdrawn.

1 If (i)
θt is negative then it corresponds to the number of units sold short.
A Brief Review of Derivatives Pricing & Hedging 3

Exercise 2 Show that if a trading strategy, θt , is s.f. then the corresponding value process, Vt , satisfies
N  
(i) (i) (i)
X
Vt+1 − Vt = θt+1 St+1 − St . (1)
i=0

Exercise 2 states that the changes in the value of the portfolio (that follows a s.f. trading strategy) are due to
capital gains or losses and are not due to the injection or withdrawal of funds. Note that we can also write (1) as

dVt = θt> dSt ,

which anticipates the continuous-time definition of an s.f. trading strategy. We can now the concepts of
arbitrage, attainable claims and completeness.

Arbitrage
Definition 5 We define a type A arbitrage opportunity to be a self-financing trading strategy, θt , such that
V0 (θ) < 0 and VT (θ) = 0. Similarly, a type B arbitrage opportunity is defined to be a self-financing trading
strategy, θt , such that V0 (θ) = 0, VT (θ) ≥ 0 and EP
0 [VT (θ)] > 0.

Attainability and Complete Markets

Definition 6 A contingent claim, C, is a random variable whose value at time T is known at that time
given the information available then. It can be interpreted as the time T value of a security.

Definition 7 We say that the contingent claim C is attainable if there exists a self-financing trading
strategy, θt , whose value process, VT , satisfies VT = C.

Note that the value of the claim, C, in Definition 7 must equal the initial value of the replicating portfolio, V0 , if
there are no arbitrage opportunities available. We can now define completeness.

Definition 8 We say that the market is complete if every contingent claim is attainable. Otherwise the
market is said to be incomplete.

We also need the definition of a numeraire:

Definition 9 A numeraire security is a security with a strictly positive price at all times, t.

It is often convenient to express the price of a security in units of a chosen numeraire. For example, if the nth
security is the numeraire security, then we define
(i)
(i) St (ωj )
S t (ωj ) := (n)
St (ωj )

to be the date t, state ωj price (in units of the numeraire security) of the ith security. We say that we are
deflating by the nth or numeraire security. Note that the deflated price of the numeraire security is always
constant and equal to 1.

Definition 10 The cash account is a particular security that earns interest at the risk-free rate of interest. In
a single period model, the date t = 1 value of the cash account is 1 + r (assuming that $1 had been deposited at
date t = 0), regardless of the terminal state and where r is the one-period interest rate that prevailed at t = 0.

In practice, we often deflate by the cash account if it exists. Note that deflating by the cash account is then
(0)
equivalent to the usual process of discounting. We will use the zeroth security with price process, St , to
denote the cash account whenever it is available. With our definitions of a numeraire security and the cash
account remaining unchanged, we can now define what we mean by an equivalent martingale measure
(EMM), or set of risk-neutral probabilities.
A Brief Review of Derivatives Pricing & Hedging 4

Equivalent Martingale Measures (EMMs)

(n)
We assume again that we have in mind a specific numeraire security with price process, St .
Definition 11 An equivalent martingale measure (EMM), Q = (q1 , . . . , qm ), is a set of probabilities
such that
1. qi > 0 for all i = 1, . . . , m.
2. The deflated security prices are martingales. That is
" (i)
#
(i)
(i) St St+s h (i) i
St := (n)
= EQ
t (n)
=: EQ
t S t+s
St St+s
for s, t ≥ 0, for all i = 0, . . . , N , and where EQ
t [.] denotes the expectation under Q conditional on
information available at time t. (We also refer to Q as a set of risk-neutral probabilities.)

1.1 Absence of Arbitrage ≡ Existence of EMM

We begin with two important propositions.

Proposition 1 If an equivalent martingale measure, Q, exists, then the deflated value process, Vt , of any
self-financing trading strategy is a Q-martingale.
(n)
Proof: Let θt be the self-financing trading strategy and let V t+1 := Vt+1 /St+1 denote the deflated value
process. We then have
"N #
Q  Q
X (i) (i)
Et V t+1 = Et θt+1 S t+1
i=0
N h (i) i
(i)
X
= θt+1 EQ
t S t+1
i=0
N
X (i) (i)
= θt+1 S t
i=0
= V t.
We can then apply the tower property of conditional expectations to see that EQ
 
t V t+s = V t for any s ≥ 0.
This completes the proof.

Remark 1 Note that Proposition 1 implies that the deflated price, V t , of any attainable security can be
computed as the Q-expectation of the terminal deflated value of the security.

Proposition 2 If an equivalent martingale measure, Q, exists, then there can be no arbitrage opportunities.
Proof: The proof follows almost immediately from Proposition 1.

We can now now state the principal result for multi-period models, assuming as usual that a numeraire security
exists.

Theorem 3 (Fundamental Theorem of Asset Pricing: Part 1)

In the multi-period model there is no arbitrage if and only if there exists an EMM, Q.
We will not prove Theorem 3 but note that one direction follows immediately from Proposition 2. The other
direction can be proved quite easily using linear programming (LP) duality techniques. We could, for example,
use LP duality to first prove the result for one-period models. We could then extend the result to multi-period
models by noting that each multi-period model is composed of a series of embedded one-period models and
then constructing the multi-period EMM from the various one-period EMMs etc.
A Brief Review of Derivatives Pricing & Hedging 5

1.2 Complete Markets ≡ Existence of a Unique EMM

In this subsection we assume that there is no arbitrage so that an EMM is guaranteed to exist. The following
proposition makes intuitive sense and is straightforward to prove.

Proposition 4 The market is complete if and only if every embedded one-period model is complete.

We then have the following theorem.

Theorem 5 (Fundamental Theorem of Asset Pricing: Part 2)

Assume there exists a security with strictly positive price process and that there are no arbitrage opportunities.
Then the market is complete if and only if there exists exactly one risk-neutral martingale measure, Q.

As was the case with Theorem 3 we could prove Theorem 5 by first proving it for one-period models and then
building up to multi-period models in a straightforward manner.

1.3 Dividends and Intermediate Cash-Flows

Thus far, we have assumed that none of the securities pay intermediate cash-flows. An example of such a
security is a dividend-paying stock. This is not an issue in the single period models since any such cash-flows are
captured in the date t = 1 value of the securities. For multi-period models, however, we sometimes need to
explicitly model these intermediate cash payments. All of the results that we have derived in these notes still go
through, however, as long as we make suitable adjustments to our price processes and are careful with our
bookkeeping. In particular, deflated cumulative gains processes rather than deflated security prices are now
Q-martingales. The cumulative gain process, Gt , of a security at time t is equal to value of the security at time
t plus accumulated cash payments that result from holding the security until time t.
Consider our discrete-time, discrete-space framework where a particular security pays dividends. Then if the
model is arbitrage-free there exists an EMM, Q, such that
 
t+s
X
S t = EQt
 Dj + S t+s  (2)
j=t+1

where Dj is the time j dividend that you receive if you hold one unit of the security, and St is its time t
ex-dividend price. This result is easy to derive using our earlier results. All we have to do is view each
dividend as a separate security with St then interpreted as the price of the portfolio consisting of these individual
securities as well as a security that is worth St+s at date t + s. The definitions of complete and incomplete
markets are unchanged and the associated results we derived earlier still hold when we also account for the
dividends in the various payoff matrices. For example, if θt is a self-financing strategy in a model with dividends
then Vt , the corresponding value process, should satisfy
N  
(i) (i) (i) (i)
X
Vt+1 − Vt = θt+1 St+1 + Dt+1 − St . (3)
i=0

(i)
Note that the time t dividends, Dt , do not appear in (3) since we assume that Vt is the value of the portfolio
just after dividends have been paid. This interpretation is consistent with taking St to be the time t ex-dividend
price of the security.
The various definitions of complete and incomplete markets, state prices, arbitrage etc. are all unchanged when
securities can pay dividends. As mentioned earlier, the First Fundamental Theorem of Asset Pricing now states
that deflated cumulative gains processes rather than deflated security prices are now Q-martingales. The second
fundamental theorem goes through unchanged.
The two fundamental theorems of asset pricing are the cornerstone of derivatives pricing theory. Moreover, given
that they hold for multi-period models it should be no surprise that they hold more generally general for
continuous-time models although some additional technical assumptions are also required then. The first
A Brief Review of Derivatives Pricing & Hedging 6

derivatives pricing models, i.e. the Black-Scholes model and the (equivalent) discrete-time binomial model, were
complete-market models where all derivative securities could be dynamically replicated. It should be clear (why?)
that the dynamic replication and martingale approaches yield the same price for any derivative security that is
replicable. When a derivative security is not replicable – such securities must exist in incomplete market models
– a unique arbitrage-free price cannot be computed. (This is simply another way of saying the derivative cannot
be replicated by an s.f. trading strategy.) But the second fundamental theorem does tell us nonetheless that, in
the absence of arbitrage, there exists infinitely many EMMs that can be used to construct arbitrage-free prices.

1.4 Derivatives Pricing in Practice

Derivatives pricing in practice often proceeds as follows: security price and / or other state variable2 are
modeled directly under an EMM, Q(φ), where φ is some parameter vector. All derivative securities (replicable or
not) are then priced using (2) with Q = Q(φ). Proposition 2 then guarantees that within this model there can
be no arbitrage. A specific value of φ is chosen by calibrating the model. That is we choose φ so that the model
prices of liquid securities coincide with the prices of those securities that we can see in the market-place. In
practice we typically cannot perform a perfect calibration so instead we solve a least-squares problem. In
particular, we minimize (over φ) the (weighted) sum-of-squared differences between model and market prices of
liquid securities. The resulting calibrated model can then be used for hedging purposes or to obtain model prices
of exotic or less liquid securities.

2 The Binomial Model

The binomial model is a discrete-time, discrete space model that describes the price evolution of a single risky
stock3 that does not pay dividends. If the stock price at the beginning of a period is S then it will either
increase to uS or decrease to dS at the beginning of the next period. In the model below we have set S0 = 100,
u = 1.06 and d = 1/u.
The binomial model assumes that there is also a cash account available that earns risk-free interest at a gross
rate of R per period. We assume R is constant4 and that the two securities (stock and cash account) may be
purchased or sold short. We let Bk = Rk denote the value at time k of $1 that was invested in the cash
account at date 0.

119.1016

112.36
PPP
106   PP 106
P
 P P 
 PP 100 
100  PP 
PP PP
PP94.3396  PP94.3396
PP P
PP 
PP88.9996
PP 
PP
P83.9619
PP

t=0 t=1 t=2 t=3

2 In term-structure models, for example, we typically model the dynamics of interest-rates rather than bond prices. These

interest-rates are the state variables in these models.

3 Binomial models are often used to model commodity and foreign exchange prices, as well as dividend-paying stock prices

and interest rate dynamics.

4 This assumption may easily be relaxed.
A Brief Review of Derivatives Pricing & Hedging 7

2.1 Martingale Pricing in the Binomial Model

We have the following result which we will prove using martingale pricing. It is also possible to derive this result
using replicating5 arguments.

Proposition 6 The binomial model is arbitrage-free if and only if

d < R < u. (4)

Proof: The first fundamental theorem of asset pricing states that there is no arbitrage in any of the embedded
one-period models at time t if and only if there exists a q satisfying 0 < q < 1 such that
 
St Q St+1
= Et
Rt Rt+1
uSt dSt
= q t+1 + (1 − q) t+1 . (5)
R R
Solving (5), we find that q = (R − d)/(u − d) and 1 − q = (u − R)/(u − d). The result now follows since
each of the embedded one-period models in the binomial model are identical.

Note that the q we obtained in the above Proposition was both unique and node independent. Therefore the
binomial model itself is arbitrage-free and complete6 if (4) is satisfied and we will always assume this to be the
case. We will usually use the cash account, Bk , as the numeraire security so that the price of any security can
be computed as the discounted expected payoff of the security under Q. Thus the time t price of a security7
that is worth XT at time T (and does not provide any cash flows in between) is given by
 
XT 1
Xt = Bt EQ t = T −t EQ t [XT ]. (6)
BT R
The binomial model is one of the workhorses of financial engineering. In addition to being a complete model, it
is also recombining. For example, an up-move followed by a down-move leads to the same node as a down-move
followed by an up-move. This recombining feature implies that the number of nodes in the tree grows linearly
with the number of time periods rather than exponentially. This leads to a considerable gain in computational
efficiency when it comes to pricing path-independent securities.

Example 2 (Pricing a Call Option)

Compute the price of a European call option on the security of Figure 1 with expiration at T = 3, and strike
K = 95. Assume also that R = 1.02.
−1
Solution: First, we find q = R−d 1.02−1.06
u−d = 1.6−1.06−1 = 0.657 which is the Q-probability of an up-move. If C0 denotes
the date 0 price of the option then (6) implies that it is given by
1 Q 1
C0 = E [CT ] = 3 EQ
0 [max(0, S3 − 95)]. (7)
R3 0 R
At this point, there are two possible ways in which we can proceed:
(i) Compute the Q-probabilities of the terminal nodes and then use (7) to determine C0 . This method does not
bother to compute the intermediate prices, Ct .
(ii) Alternatively, we can work backwards in the lattice one period at a time to find Ct at each node and at
each time t. This procedure is sometimes referred to as backwards recursion and is no more than a simple
application of dynamic programming.
5 Indeed, most treatments of the binomial model use replicating arguments to show that (4) must hold in order to rule out

arbitrage.
6 We could also have argued completeness by observing that the matrix of payoffs corresponding to each embedded one-period

model has rank 2 which is equal to the number of possible outcomes.

7 X is also the time t value of a self-financing trading strategy that replicates X .
t T
A Brief Review of Derivatives Pricing & Hedging 8

Stock Price European Option Price

119.10 24.10
112.36 106.00 19.22 11.00
106.00 100.00 94.34 14.76 7.08 0.00
100.00 94.34 89.00 83.96 11.04 4.56 0.00 0.00

t=0 t=1 t=2 t=3 t=0 t=1 t=2 t=3

For example in the European Option Payoff table above, we see that 14.76 = R1 (q(19.22) + (1 − q)(7.08)),
i.e., the value of the option at any node is the discounted expected value of the option one time period ahead.
This is just restating the Q-martingale property of discounted security price processes. We find that the call
option price at t = 0 is given by $11.04.

Example 3 (A Counter-Intuitive Result)

Consider the same option-pricing problem of Example 2 except that we now take R = 1.04. We then obtain a
European call option price of 15.64 as may be seen from the lattices given below. Note that this price is greater
than the option price, 11.04, that we obtained in Example 2.

Stock Price European Option Price

119.10 24.10
112.36 106.00 21.01 11.00
106.00 100.00 94.34 18.19 8.76 0.00
100.00 94.34 89.00 83.96 15.64 6.98 0.00 0.00

t=0 t=1 t=2 t=3 t=0 t=1 t=2 t=3

This observation seems counterintuitive: after all, we are dealing only with positive cash flows, the values of
which have not changed, i.e. the option payoffs upon expiration at t = 3 have not changed. On the other hand,
the interest rate that is used to discount cash flows has increased in this example and so we might have
expected the value of the option to have decreased. What has happened? (Certainly this situation would never
have occurred in a deterministic world!)
First, from a purely mechanical viewpoint we can see that the risk-neutral probabilities have changed. In
particular, the risk-neutral probability of an up-move, q = (R − d)/(u − d), has increased since R has increased.
This means that we are more likely to end up in the higher-payoff states. This increased likelihood of higher
payoffs more than offsets the cost of having a larger discount factor and so we ultimately obtain an increase in
the option value.
This, however, is only one aspect of the explanation. It is perhaps more interesting to look for an intuitive
explanation as to why q should increase when R increases. You should think about this!

2.2 Calibrating the Binomial Model

In continuous-time models, it is often assumed that security price processes are geometric Brownian motions. In
that case we write St ∼ GBM(µ, σ) if
2
/2)s + σ(Bt+s −Bt )
St+s = St e(µ−σ (8)
A Brief Review of Derivatives Pricing & Hedging 9

where Bt is a standard Brownian motion. Note that this model has the nice property that the gross return,
Rt,t+s , in any period, [t, t + s], is independent of returns in earlier periods. In particular, it is independent of St .
This follows by noting
St+s 2
Rt,t+s = = e(µ−σ /2)s + σ(Bt+s −Bt )
St
and recalling the independent increments property of Brownian motion. It is appealing8 that Rt,t+s is
independent of St since it models real world markets where investors care only about returns and not the
absolute price level of securities. The binomial model has similar properties since the gross return in any period
of the binomial model is either u or d, and this is independent of what has happened in earlier periods.
We often wish to calibrate the binomial model so that its dynamics match that of the geometric Brownian
motion in (8). To do this we need to choose u, d and p, the real-world probability of an up-move, appropriately.
There are many possible ways of doing this, but one of the more common choices9 is to set

eµT /n − dn
pn = (9)
un − dn
p
un = exp(σ T /n) (10)
p
dn = 1/un = exp(−σ T /n) (11)

where T is the expiration date and n is the number of periods. (This calibration becomes more accurate as n
increases.) Note then, for example, that E[Si+1 |Si ] = pn un Si + (1 − pn )dn Si = Si exp(µT /n), as desired.
We will choose the gross risk-free rate per period, Rn , so that it corresponds to a continuously-compounded
rate, r, in continuous time. We therefore have

Rn = erT /n . (12)

Remark 2 Recall that the true probability of an up-move, p, has no bearing upon the risk-neutral probability,
q, and therefore it does not directly affect how securities are priced. From our calibration of the binomial model,
we therefore see that µ, which enters the calibration only through p, does not impact security prices. On the
other hand, u and d depend on σ which therefore does impact security prices. This is a recurring theme in
derivatives pricing.

Remark 3 We just stated that p does not directly affect how securities are priced. This means that if p should
suddenly change but S0 , R, u and d remain unchanged, then q, and therefore derivative prices, would also
remain unchanged. This seems very counter-intuitive but an explanation is easily given. In practice, a change in
p would generally cause one or more of S0 , R, u and d to also change. This would in turn cause q, and therefore
derivative prices, to change. We could therefore say that p has an indirect effect on derivative security prices.

2.3 Convergence of the Binomial Model to Black-Scholes

The Black-Scholes formula for the price of a call option on a non-dividend paying security with initial price S0 ,
strike K, time to expiration T , continuously compounded interest rate r, and volatility parameter σ, is given by

C(S0 , T ) = S0 N (d1 ) − Ke−rT N (d2 ) (13)

where
log(S0 /K) + (r + σ 2 /2)T
d1 = √ ,
σ T
√
d2 = d1 − σ T
8 More sophisticated models will sometimes allow for return predictability where Rt,t+s is not independent of St . Even then,
it is still appropriate to model returns rather than absolute security values.
9 We write p , u and d to emphasize that their values depend explicitly on the number of periods, n, for a fixed expiration,
n n n
T.
A Brief Review of Derivatives Pricing & Hedging 10

and N (·) is the CDF of a standard Normal random variable.

One way to obtain the Black-Scholes formula is to consider the sequence of binomial models, Mn , that are
parameterized by (9), (10), (11) and (12). Letting n → ∞ in this sequence it can be shown via a version of the
Central Limit Theorem that the call option price in the binomial model will converge to (13).

3 The Volatility Surface

The Black-Scholes model is an elegant model but it does not perform very well in practice. For example, it is
well known that stock prices jump on occasions and do not always move in the smooth manner predicted by the
GBM motion model. Stock prices also tend to have fatter tails than those predicted by GBM. Finally, if the
Black-Scholes model were correct then we should have a flat implied volatility surface. The volatility surface is a
function of strike, K, and time-to-maturity, T , and is defined implicitly

C(S, K, T ) := BS (S, T, r, q, K, σ(K, T )) (14)

where C(S, K, T ) denotes the current market price of a call option with time-to-maturity T and strike K, and
BS(·) is the Black-Scholes formula for pricing a call option. In other words, σ(K, T ) is the volatility that, when
substituted into the Black-Scholes formula, gives the market price, C(S, K, T ). Because the Black-Scholes
formula is continuous and increasing in σ, there will always10 be a unique solution, σ(K, T ). If the
Black-Scholes model were correct then the volatility surface would be flat with σ(K, T ) = σ for all K and T . In
practice, however, not only is the volatility surface not flat but it actually varies, often significantly, with time.

Figure 1: The Volatility Surface

In Figure 1 above we see a snapshot of the11 volatility surface for the Eurostoxx 50 index on November 28th ,
2007. The principal features of the volatility surface is that options with lower strikes tend to have higher
implied volatilities. For a given maturity, T , this feature is typically referred to as the volatility skew or smile.
For a given strike, K, the implied volatility can be either increasing or decreasing with time-to-maturity. In
general, however, σ(K, T ) tends to converge to a constant as T → ∞. For T small, however, we often observe
10 Assuming there is no arbitrage in the market-place.
11 Note that by put-call parity the implied volatility σ(K, T ) for a given European call option will be also be the implied
volatility for a European put option of the same strike and maturity. Hence we can talk about “the” implied volatility surface.
A Brief Review of Derivatives Pricing & Hedging 11

an inverted volatility surface with short-term options having much higher volatilities than longer-term options.
This is particularly true in times of market stress.
It is worth pointing out that different implementations12 of Black-Scholes will result in different implied volatility
surfaces. If the implementations are correct, however, then we would expect the volatility surfaces to be very
similar in shape. Single-stock options are generally American and in this case, put and call options will typically
give rise to different surfaces. Note that put-call parity does not apply for American options.
Clearly then the Black-Scholes model is far from accurate and market participants are well aware of this.
However, the language of Black-Scholes is pervasive. Every trading desk computes the Black-Scholes implied
volatility surface and the Greeks they compute and use are Black-Scholes Greeks.

Arbitrage Constraints on the Volatility Surface

The shape of the implied volatility surface is constrained by the absence of arbitrage. In particular:
1. We must have σ(K, T ) ≥ 0 for all strikes K and expirations T .
2. At any given maturity, T , the skew cannot be too steep. Otherwise butterfly arbitrages will exist. For
example fix a maturity, T and consider put two options with strikes K1 < K2 . If there is no arbitrage then
it must be the case (why?) that P (K1 ) < P (K2 ) where P (Ki ) is the price of the put option with strike
Ki . However, if the skew is too steep then we would obtain (why?) P (K1 ) > P (K2 ).
3. Likewise the term structure of implied volatility cannot be too inverted. Otherwise calendar spread
arbitrages will exist. This is most easily seen in the case where r = q = 0. Then, fixing a strike K, we can
let Ct (T ) denote the time t price of a call option with strike K and maturity T . Martingale pricing implies
that Ct (T ) = Et [(ST − K)+ ]. We have seen before that (ST − K)+ is a Q-submartingale and now
standard martingale results can be used to show that Ct (T ) must be non-decreasing in T . This would be
violated (Why?) if the term structure of implied volatility was too inverted.
In practice the implied volatility surface will not violate any of these restrictions as otherwise there would be an
arbitrage in the market. These restrictions can be difficult to enforce, however, when we are “bumping” or
“stressing” the volatility surface, a task that is commonly performed for risk management purposes.

Why is there a Skew?

For stocks and stock indices the shape of the volatility surface is always changing. There is generally a skew,
however, so that for any fixed maturity, T , the implied volatility decreases with the strike, K. It is most
pronounced at shorter expirations. There are two principal explanations for the skew.
1. Risk aversion which can appear as an explanation in many guises:
(a) Stocks do not follow GBM with a fixed volatility. Markets often jump and jumps to the downside
tend to be larger and more frequent than jumps to the upside.
(b) As markets go down, fear sets in and volatility goes up.
(c) Supply and demand. Investors like to protect their portfolio by purchasing out-of-the-money puts and
so there is more demand for options with lower strikes.
2. The leverage effect which is due to the fact that the total value of company assets, i.e. debt + equity,
is a more natural candidate to follow GBM. If so, then equity volatility should increase as the equity value
decreases. To see this consider the following:

Let V , E and D denote the total value of a company, the company’s equity and the company’s debt,
respectively. Then the fundamental accounting equations states that

V = D + E. (15)
12 For example different methods of handling dividends would result in different implementations.
A Brief Review of Derivatives Pricing & Hedging 12

Equation (15) is the basis for the classical structural models that are used to price risky debt and credit
default swaps. Merton (1970’s) recognized that the equity value could be viewed as the value of a call
option on V with strike equal to D.

Let ∆V , ∆E and ∆D be the change in values of V , E and D, respectively. Then

V + ∆V = (E + ∆E) + (D + ∆D) so that
V + ∆V E + ∆E D + ∆D
= +
V V V  
E E + ∆E D D + ∆D
= + (16)
V E V D

If the equity component is substantial so that the debt is not too risky, then (16) implies
E
σV ≈ σE
V
where σV and σE are the firm value and equity volatilities, respectively. We therefore have
V
σE ≈ σV . (17)
E

Example 4 (The Leverage Effect)

Suppose, for example, that V = 1, E = .5 and σV = 20%. Then (17) implies σE ≈ 40%. Suppose σV remains
unchanged but that over time the firm loses 20% of its value. Almost all of this loss is borne by equity so that
now (17) implies σE ≈ 53%. σE has therefore increased despite the fact that σV has remained constant.

It is interesting to note that there was little or no skew in the market before the Wall street crash of 1987. So it
appears to be the case that it took the market the best part of two decades before it understood that it was
pricing options incorrectly.

What the Volatility Surface Tells Us

To be clear, we continue to assume that the volatility surface has been constructed from European option prices.
Consider a butterfly strategy centered at K where you are:
1. long a call option with strike K − ∆K
2. long a call with strike K + ∆K
3. short 2 call options with strike K
The value of the butterfly, B0 , at time t = 0, satisfies

B0 = C(K − ∆K, T ) − 2C(K, T ) + C(K + ∆K, T )

≈ e−rT Prob(K − ∆K ≤ ST ≤ K + ∆K) × ∆K/2
≈ e−rT f (K, T ) × 2∆K × ∆K/2
= e−rT f (K, T ) × (∆K)2

where f (K, T ) is the probability density function (PDF) of ST evaluated at K. We therefore have

C(K − ∆K, T ) − 2C(K, T ) + C(K + ∆K, T )

f (K, T ) ≈ erT . (18)
(∆K)2

Letting ∆K → 0 in (18), we obtain

∂2C
f (K, T ) = erT .
∂K 2
A Brief Review of Derivatives Pricing & Hedging 13

The volatility surface therefore gives the marginal risk-neutral distribution of the stock price, ST , for any time,
T . It tells us nothing about the joint distribution of the stock price at multiple times, T1 , . . . , Tn .
This should not be surprising since the volatility surface is constructed from European option prices and the
latter only depend on the marginal distributions of ST .

Example 5 (Same marginals, different joint distributions)

Suppose there are two time periods, T1 and T2 , of interest and that a non-dividend paying security has
risk-neutral distributions given by
2 √
ST1 = e(r−σ /2)T1 +σ T1 Z1
(19)
√ √

(r−σ 2 /2)T2 +σ T2 ρZ1 + 1−ρ2 Z2
ST2 = e (20)

where Z1 and Z2 are independent N(0, 1) random variables. Note that a value of ρ > 0 can capture a
momentum effect and a value of ρ < 0 can capture a mean-reversion effect. We are also implicitly assuming
that S0 = 1.
(A) (B)
Suppose now that there are two securities, A and B say, with prices St and St given by (19) and (20) at
times t = T1 and t = T2 , and with parameters ρ = ρA and ρ = ρB , respectively. Note that the marginal
(A) (B)
distribution of St is identical to the marginal distribution of St for t ∈ {T1 , T2 }. It therefore follows that
options on A and B with the same strike and maturity must have the same price. A and B therefore have
identical volatility surfaces.
But now consider a knock-in put option with strike 1 and expiration T2 . In order to knock-in, the stock price at
time T1 must exceed the barrier price of 1.2. The payoff function is then given by

Payoff = max (1 − ST2 , 0) 1{ST1 ≥1.2} .

Question: Would the knock-in put option on A have the same price as the knock-in put option on B?
Question: How does your answer depend on ρA and ρB ?
Question: What does this say about the ability of the volatility surface to price barrier options?

4 The Greeks
We now turn to the sensitivities of the option prices to the various parameters. These sensitivities, or the
Greeks are usually computed using the Black-Scholes formula, despite the fact that the Black-Scholes model is
known to be a poor approximation to reality. But first we return to put-call parity.

Put-Call Parity
Consider a European call option and a European put option, respectively, each with the same strike, K, and
maturity T . Assuming a continuous dividend yield, q, then put-call parity states

e−rT K + Call Price = e−qT S + Put Price. (21)

This of course follows from a simple arbitrage argument and the fact that both sides of (21) equal max(ST , K)
at time T . Put-call parity is useful for calculating Greeks. For example13 , it implies that
Vega(Call) = Vega(Put) and that Gamma(Call) = Gamma(Put). It is also extremely useful for calibrating
13 See below for definitions of vega and gamma.
A Brief Review of Derivatives Pricing & Hedging 14

dividends and constructing the volatility surface.

The Greeks
The principal Greeks for European call options are described below. The Greeks for put options can be
calculated in the same manner or via put-call parity.
Definition: The delta of an option is the sensitivity of the option price to a change in the price of the
underlying security.

The delta of a European call option satisfies

∂C
delta = = e−qT Φ(d1 ).
∂S
This is the usual delta corresponding to a volatility surface that is sticky-by-strike. It assumes that as the
underlying security moves, the volatility of the option does not move. If the volatility of the option did move
then the delta would have an additional term of the form vega × ∂σ(K, T )/∂S.

(a) Delta for European Call and Put Options (b) Delta for Call Options as Time-To-Maturity Varies

Figure 2: Delta for European Options

By put-call parity, we have deltaput = deltacall − e−qT . Figure 2(a) shows the delta for a call and put option,
respectively, as a function of the underlying stock price. In Figure 2(b) we show the delta for a call option as a
function of the underlying stock price for three different times-to-maturity. It was assumed r = q = 0. What is
the strike K? Note that the delta becomes steeper around K when time-to-maturity decreases. Note also that
delta = Φ(d1 ) = Prob(option expires in the money). (This is only approximately true when r and q are
non-zero.)
In Figure 3 we show the delta of a call option as a function of time-to-maturity for three options of different
money-ness. Are there any surprises here? What would the corresponding plot for put options look like?

Definition: The gamma of an option is the sensitivity of the option’s delta to a change in the price of the
underlying security.

The gamma of a call option satisfies

∂2C φ(d1 )
gamma = = e−qT √
∂S 2 σS T
A Brief Review of Derivatives Pricing & Hedging 15

Figure 3: Delta for European Call Options as a Function of Time-To-Maturity

where φ(·) is the standard normal PDF.

(a) Gamma as a Function of Stock Price (b) Gamma as a Function of Time-to-Maturity

Figure 4: Gamma for European Options

In Figure 4(a) we show the gamma of a European option as a function of stock price for three different
time-to-maturities. Note that by put-call parity, the gamma for European call and put options with the same
strike are equal. Gamma is always positive due to option convexity. Traders who are long gamma can make
money by gamma scalping. Gamma scalping is the process of regularly re-balancing your options portfolio to be
delta-neutral. However, you must pay for this long gamma position up front with the option premium. In Figure
4(b), we display gamma as a function of time-to-maturity. Can you explain the behavior of the three curves in
Figure 4(b)?

Definition: The vega of an option is the sensitivity of the option price to a change in volatility.

The vega of a call option satisfies

∂C √
vega = = e−qT S T φ(d1 ).
∂σ
A Brief Review of Derivatives Pricing & Hedging 16

(a) Vega as a Function of Stock Price (b) Vega as a Function of Time-to-Maturity

Figure 5: Vega for European Options

In Figure 5(b) we plot vega as a function of the underlying stock price. We assumed K = 100 and that
r = q = 0. Note again that by put-call parity, the vega of a call option equals the vega of a put option with the
same strike. Why does vega increase with time-to-maturity? For a given time-to-maturity, why is vega peaked
near the strike? Turning to Figure 5(b), note that the vega decreases to 0 as time-to-maturity goes to 0. This is
consistent with Figure 5(a). It is also clear from the expression for vega.

Question: Is there any “inconsistency” to talk about vega when we use the Black-Scholes model?

Definition: The theta of an option is the sensitivity of the option price to a negative change in
time-to-maturity.

The theta of a call option satisfies

∂C σ
theta = − = −e−qT Sφ(d1 ) √ + qe−qT SN(d1 ) − rKe−rT N(d2 ).
∂T 2 T

In Figure 6(a) we plot theta for three call options of different times-to-maturity as a function of the underlying
stock price. We have assumed that r = q = 0%. Note that the call option’s theta is always negative. Can you
explain why this is the case? Why does theta become more negatively peaked as time-to-maturity decreases to
0?
In Figure 6(b) we again plot theta for three call options of different money-ness, but this time as a function of
time-to-maturity. Note that the ATM option has the most negative theta and this gets more negative as
time-to-maturity goes to 0. Can you explain why?

Options Can Have Positive Theta: In Figure 7 we plot theta for three put options of different money-ness
as a function of time-to-maturity. We assume here that q = 0 and r = 10%. Note that theta can be positive for
in-the-money put options. Why? We can also obtain positive theta for call options when q is large. In typical
scenarios, however, theta for both call and put options will be negative.
A Brief Review of Derivatives Pricing & Hedging 17

(a) Theta as a Function of Stock Price (b) Theta as a Function of Time-to-Maturity

Figure 6: Theta for European Options

Figure 7: Positive Theta is Possible

The Relationship between Delta, Theta and Gamma

Recall that the Black-Scholes PDE states that any derivative security with price Pt must satisfy

∂P ∂P 1 ∂2P
+ (r − q)S + σ 2 S 2 2 = rP. (22)
∂t ∂S 2 ∂S
Writing θ, δ and Γ for theta, delta and gamma, we obtain
1
θ + (r − q)Sδ + σ 2 S 2 Γ = rP. (23)
2
Equation (23) holds in general for any portfolio of securities. If the portfolio in question is delta-hedged so that
the portfolio δ = 0 then we obtain
1
θ + σ 2 S 2 Γ = rP (24)
2
It is clear from (24) that any gain from gamma is offset by losses due to theta. This of course assumes that the
correct implied volatility is assumed in the Black-Scholes model. Since we know that the Black-Scholes model is
A Brief Review of Derivatives Pricing & Hedging 18

wrong, this observation should only be used to help your intuition and not taken as a “fact”.

Delta-Gamma-Vega Approximations to Option Prices

A simple application of Taylor’s Theorem says

∂C 1 ∂2C ∂C
C(S + ∆S, σ + ∆σ) ≈ C(S, σ) + ∆S + (∆S)2 2 + ∆σ
∂S 2 ∂S ∂σ
1
= C(S, σ) + ∆S × δ + (∆S)2 × Γ + ∆σ × vega.
2
where C(S, σ) is the price of a derivative security as a function14 of the current stock price, S, and the implied
volatility, σ. We therefore obtain
Γ
P&L = (∆S)2 + vega ∆σ
δ∆S +
2
= delta P&L + gamma P&L + vega P&L

When ∆σ = 0, we obtain the well-known delta-gamma approximation. This approximation is often used, for
example,in historical Value-at-Risk (VaR) calculations for portfolios that include options. We can also write
2
ΓS 2
  
∆S ∆S
P&L = δS + + vega ∆σ
S 2 S
= ESP × Return + $ Gamma × Return2 + vega ∆σ

where ESP denotes the equivalent stock position or “dollar” delta.

5 Delta Hedging
In the Black-Scholes model with GBM, an option can be replicated15 exactly via delta-hedging. The idea
behind delta-hedging is to continuously re-balance an s.f. portfolio that continuously trades the underlying stock
and cash account so that the portfolio always has a delta equal to the delta of the option being hedged. Of
course in practice we cannot hedge continuously and so instead we hedge periodically. Periodic or discrete
hedging then results in some replication error.
Delta-hedging proceeds as follows. The portfolio begins with an initial cash value of C0 which is the time t = 0
value of the position or option we want to hedge. Then, and in each subsequent period, the portfolio is
re-balanced in such a way that at time t we are long δt units of the stock where δt is the time t delta of the
option. Any remaining cash is invested in the cash account and therefore earns at the risk-free rate of interest.
The stock position accrues dividends according to the dividend yield, q. Let Pt denote the time t value of this
discrete-time trading strategy. The value of the corresponding portfolio then evolves according to

P0 := C0 (25)


Pti+1 = Pti + (Pti − δti Sti ) r∆t + δti Sti+1 − Sti + qSti ∆t (26)

where ∆t := ti+1 − ti which we assume is constant for all i and r is the annual risk-free interest rate (assuming
per-period compounding). Note that δti is a function of Sti and some assumed implied volatility, σimp say. We
also note that (25) and (26) respect the self-financing condition. Stock prices are simulated assuming
St ∼ GBM(µ, σ) so that √
2
St+∆t = St e(µ−σ /2)∆t+σ ∆tZ
14 Theprice may also depend on other parameters, in particular time-to-maturity, but we suppress that dependence here.
15 In
fact one way to obtain the Black-Scholes formula is via a replication argument that leads to the so-called Black-Scholes
PDE. The solution of this PDE is the Black-Scholes formula.
A Brief Review of Derivatives Pricing & Hedging 19

where Z ∼ N(0, 1). Note the option implied volatility, σimp , need not equal σ which in turn need not equal the
realized volatility (when we hedge periodically as opposed to continuously). This has interesting implications for
the trading P&L which we may define as
P&L := PT − CT
= PT − (ST − K)+
in the case of a short position in a call option with strike K and maturity T . Note that PT is the terminal value
of the replicating strategy in (26). Many interesting questions now arise:

Question: If you sell options, what typically happens the total P&L if σ < σimp ?
Question: If you sell options, what typically happens the total P&L if σ > σimp ?
Question: If σ = σimp what typically happens the total P&L as the number of re-balances increases?

Some Answers to Delta-Hedging Questions

Recall that the price of an option increases as the volatility increases. Therefore if realized volatility is higher
than expected, i.e. the level at which it was sold, we expect to lose money on average when we delta-hedge an
option that we sold. Similarly, we expect to make money when we delta-hedge if the realized volatility is lower
than the level at which it was sold.
In general, however, the payoff from delta-hedging an option is path-dependent, i.e. it depends on the price
path taken by the stock over the entire time interval. In fact, we can show that the payoff from continuously
delta-hedging an option (where the underlying St never jumps) satisfies
Z T 2 2
St ∂ Vt 2
− σt2 dt


P&L = 2
σimp (27)
0 2 ∂S
where Vt is the time t value of the option and σt is the realized instantaneous volatility at time t.
S2 2
The term 2t ∂∂SV2t is often called the dollar gamma, as discussed earlier. It is always positive for a call or put
option, but it goes to zero as the option moves significantly into or out of the money.

Returning to self-financing trading strategy of (25) and (26), note that we can choose any model we like for the
security price dynamics. In particular, we are not restricted to choosing geometric Brownian motion and other
diffusion or jump-diffusion models could be used instead. It is interesting to simulate these alternative models
and to then observe what happens to the replication error in (27) where the δti ’s are computed assuming
(incorrectly) a geometric Brownian motion price dynamics. This is actually the situation in practice – we don’t
know the true market dynamics but we often hedge using the Black-Scholes model.

6 Extensions of Black-Scholes
The Black-Scholes model is easily applied to other securities. In addition to options on stocks and indices, these
securities include currency options, options on some commodities and options on index, stock and currency
futures. Of course, in all of these cases it is well understood that the model has many weaknesses. As a result,
the model has been extended in many ways. These extensions include jump-diffusion models, stochastic
volatility models, local volatility models, regime-switching models, garch models and others. Most of these
models are incomplete models and the pricing philosophy then proceeds as described in Section 1.4.
One of the principal uses of the Black-Scholes framework is that is often used to quote derivatives prices via
implied volatilities. This is true even for securities where the GBM model is clearly inappropriate. Such securities
include, for example, caplets and swaptions in the fixed income markets, CDS options in credit markets and
options on variance-swaps in equity markets.