Interest Rate Models 2

Interest Rate Models

Goals:
• Describe the most common interest rate derivatives and the main
approaches to modeling them.
• Describe volatility smiles in interest rate derivatives and the most
common SDEs used.

References: Book Chapter 13

MF 728: Fixed Income

Interest Rate Models 3

Review: Interest Rate Products

• As we saw in the last lecture, there are many linear products that
we can trade, including:
– FRA’s
– SOFR Futures
– Swaps
– Basis Swaps
• We also saw that we can express each of these products in terms
of forward rates and saw how we can parameterize the yield
curve by these forward rates.

MF 728: Fixed Income

Interest Rate Models 4

Review: Approaches to Yield Curve

Construction

• Recall from the last lecture that constructing a yield curve

involves choosing the forward rates that best match market data.
• In particular, we discussed two approaches for handling this:
– Bootstrapping: Iterate through the instruments from shortest
to longest maturity fixing their forward rates as you proceed.
– Optimization: Minimize the least squares distance between
model and market rates via an optimization procedure.
• Optimization provided us with more flexibility to ensure a smooth,
intuitive curve shape but adds complexity.

MF 728: Fixed Income

Interest Rate Models 5

Review: Using Bootstrapping to Extract a

Yield Curve
• We start with the current SOFR rate to set the spot rate.
• Next we iterate through the SOFR futures. Each SOFR future is 3
months so at each point you will be setting the forward rate for a
3 month period.
• Finally, proceed to swaps beginning with the one or two-year swap
and follow the same process.
• An astute observer might notice that in some cases the SOFR
futures and short tenor swap rates span the same sections of the
forward curve.
• In these cases, we cannot match both and must either use the
SOFR futures and omit the one or two-year swap, or vice versa.

MF 728: Fixed Income

Interest Rate Models 6

Review: Using Optimization to Extract a

Yield Curve

• When using optimization we would begin with a parameterization

of the forward rate process of the curve, f (t).
• We could make f (t) piecewise constant, or use spline/b-spline
functions.
• Next, we choose an objective function that minimizes the distance
between market and model interest rates.
m
X
Q(f ) = (Rj − R̂j )2 (1)
j=1

• Recall that we could also add a regularization term in order to try

to control the jaggedness of the forward curve.

MF 728: Fixed Income

Interest Rate Models 7

Empirical Observations of Yield Curves

• Generally speaking a yield curve can be described by:
– The level of interest rates
– The slope of the curve
– The curvature of the curve
• Slope is of particular interest and is watched by market
participants.
• Upward sloping curves correspond to normal conditions with term
premium in the curve and/or expected future rate increases.
• Inverted yield curves correspond to stressed market condition and
usually signal market expectations of future rate cuts.
• The Fed and other researchers have found that often an inverted
yield curve portends a recession.

MF 728: Fixed Income

Interest Rate Models 8

Principal Component Analysis

• We can gain insight into the empirical behavior of yield curves by

examining the structure of the comovements C of different rates.
• One way to do this is via Principal Component Analysis
• PCA is at heart a matrix decomposition that creates orthogonal
factors.
• The underlying spectral decomposition of C can be written as:

C = U DU ⊤ (2)

• PCA can help us reduce dimensionality of the movements of the

yield curve and show which factors are driving the largest portions
of the variance.

MF 728: Fixed Income

Interest Rate Models 9

Yield Curve Construction: PCA of the Yield

Curve

• PCA can help determine what drives changes in the yield curve.
• This can be done by performing PCA on a covariance matrix or
correlation of empirical rate shifts.
• In practice, when we do this we generally find that the first few
principal components have intuitive interpretation and roughly
correspond to changes in:
– Level of Rates (i.e. Parallel Shifts)
– Slope of the Curve (i.e. Steepening / Flattening)
– Curvature
• These first 2-3 components generally explain ≈ 95% of the
variance of the yield curve.

MF 728: Fixed Income

Interest Rate Models 10

Interest Rate Derivatives: Overview

• Vanilla Options
– SOFR Future Options
– Caps
– Swaptions
• Exotic Options
– Bermudan Swaptions
– Spread Options
– CMS Options

MF 728: Fixed Income

Interest Rate Models 11

Linear vs. Non-Linear Payoffs

• Linear Payoffs (e.g. Swaps)

Can be replicated statically
Do not require use of stochastic models
• Vanilla Non-Linear Payoffs (e.g. Swaptions)
Liquid market pricing exists
Requires calibration of a stochastic model
Can be valued via a simple, market standard model
Requires dynamic replication
• Exotic Non-Linear Payoffs (e.g. Bermudan Swaptions)
Illiquid products with no standardized market prices.
No market standard model. Generally requires bigger, term
structure model.

MF 728: Fixed Income

Interest Rate Models 12

Numeraires in Interest Rate Modelling

• In traditional valuation of derivatives in most asset classes, we

operate in the risk neutral measure.
• Consider the following risk neutral pricing equation in the risk
neutral measure:
Z +∞ R
− 0T ru du
p0 = e F (ST )ϕ(ST ) dST (3)
−∞

• When doing this, we often assume that interest rates are

deterministic (and sometimes constant)
• In interest rate markets, however, interest rates play a role of
discounting and are also in the payoff.
• This adds another level of complexity and often means that the
risk neutral measure in (3) will not be the most convenient.

MF 728: Fixed Income

Interest Rate Models 13

What is a Numeraire?

• In this class we don’t delve too deply into the derivations of

numeraires but it is important to understand the basic
concept/intuition.
• A numeraire is a tradable asset with a strictly positive prices
process at all times.
• When solving options pricing problems, we compute the relative
price of an asset or derivative discounted by the numeraire:
S(t)
S N (t) = (4)
N (t)

• In standard pricing problems in other asset classes, the numeraire

is a zero-coupon bond, e.g.: N (t) = exp(rt).

MF 728: Fixed Income

Interest Rate Models 14

What is a Numeraire?

• While this is a convenient numeraire for equities and other asset

classes, it is not convenient in rates modelling.
• Note that the choice of a numeraire is a matter of convenience.
We will see in rates modelling this varies by the derivative we are
working with.
• This means we can work in a so called equivalent martingale
measure which may be different for swaptions than for caps.

MF 728: Fixed Income

Interest Rate Models 15

Change of Numeraire

• Arbitrage free pricing dictates that the following pricing equation

holds:
 
V (s) Q V (T )
=E (5)
N (s) N (T )

• Of course we could choose another numeraire, and Girsanov’s

theorem tells us this will only change the drift term.
• It is most convenient for us not to model the drift at all, this is
why we choose to work in a martinagle measure when we can.

MF 728: Fixed Income

Interest Rate Models 16

Change of Numeraire

• Returning to (5):
– N (s) is today’s value of the chosen numeraire.
– N (T ) is the value of the chosen numeraire at the end. It will
be convenient to choose numeraires where this term is equal to
1 (e.g., a Zero-Coupon Bond)
• Assuming we choose such a numeraire, that is N (T ) = 1, we are
left with:

V (s) = N (s)EQ [V (T )] (6)

MF 728: Fixed Income

Interest Rate Models 17

Common Numeraires in Rates Modeling

• Spot Numeraire: A deposit in a bank accruing the instantaneous

risk-free rates.
Z T !
N (T ) = exp r(s)ds (7)
0

• Note that if we assume r(s) is a constant then this corresponds to

the familiar risk neutral measure we see in equity derivatives.
• T Forward Numeraire: The price of a zero coupon bond at time
t that matures at time T.
• This arises naturally when pricing instruments with maturity T.
• The spot measure is a special case of the T Forward measure with
t = 0.

MF 728: Fixed Income

Interest Rate Models 18

Common Numeraires in Rates Modeling

• Annuity Numeraire: The value of a one basis point stream of

payments beginning at time t and ending at time T.
N
X
A(t, T ) = δti D(t, ti ) (8)
i=1

• Recall that this annuity function plays a special role in

determining the par swap rate.

MF 728: Fixed Income

Interest Rate Models 19

Simple Example: Caplets

• The simplest interest rate derivative is a caplet, which is an option

on a single forward rate.
• We can model a caplet by assuming that its underlying forward
rate follows some stochastic process.
• For example we might assume that the forward follows a Normal
or Lognormal process.
• In order to incorporate caplet volatility skew, we may need a more
realistic model, such as SABR.

MF 728: Fixed Income

Interest Rate Models 20

Caplet Pricing Formulas

• The price of a caplet in the T-Forward measure can be written as:

+
 
c(K) = δP (0, T + δ)E (FT − K) (9)

where δ is the length of the accrual period and P (0, T + δ) is the

discount factor from T + δ to the present.
• If we use Black’s model, then there is a closed-form solution to
the price of a caplet:

c = δP (0, T + δ) [F0 N (d1 ) − KN (d2 )] (10)

F0

1 2
log K + 2 σ T
d1 = √ (11)
σ T
F0

1 2
log K − 2 σ T
d2 = √ (12)
σ T

MF 728: Fixed Income

Interest Rate Models 21

Caplet Pricing Formulas: Black’s Model

Implied Volatility

• Black’s formula tells us how to compute an option price given

inputs T and σ.
• However in practice we observe bid and offer prices, and need to
extract the value of σimplied that matches the market price.
• This procedure of fitting an implied volatility to a market price
requires a one-dimensional root-finding algorithm.
• Many modern programming languages such as Python have
built-in functions for these types of calculations. In Python we can
use the optimize.root function in the scipy module.
• A similar procedure would exist for the Bachelier model, which is
also commonly applied in interest rate markets.

MF 728: Fixed Income

Interest Rate Models 22

Caplets vs Caps

• Caps are baskets of caplets consisting of multiple options each

linked to a different forward rate.
• Caps are OTC instruments.
• Caps are a function of strike, starting expiry, and length.
• As caps are baskets of caplets, their pricing equation can be
written as:
X
+
 
C(K) = δP (0, Ti + δ)E (FTi − K) (13)
i

where i indicates that we are looping over all individual caplets.

MF 728: Fixed Income

Interest Rate Models 23

Caps, Caplets & Stripping Caplet Volatilities

• In practice, the market trades caps rather than caplets and caps
are quoted in terms of a single (Black) implied volatility.
• This single implied volatility is interpreted as the constant volatility
that allows us to match the price of the entire string of caplets.
• In order to obtain volatilities from individual caplets we need to
extract them from the constant cap vols. This process is referred
to as stripping caplet volatilities.

MF 728: Fixed Income

Interest Rate Models 24

Caps, Caplets & Stripping Caplet Volatilities

• There are two main approaches to stripping caplet volatilities from

a series of cap prices or volatiliies:
– Bootstrapping: start at the shortest expiry cap and move
further away trying to match cap prices while fixing the
individual caplet volatilities.
– Optimization: start with a parameterized function for the
caplet volatilities and perform and optimization that minimizes
the pricing error over the set of parameters in the function.
• This procedure mimics the process we saw when constructing the
yield curve with a different objective function and underlying
pricing formulas.

MF 728: Fixed Income

Interest Rate Models 25

Swaptions

• Swaptions are vanilla options on LIBOR/SOFR based swaps.

• Payer swaption is the right to pay fixed and
Receiver swaption is the right to receive fixed.
• Payer (Receiver) swaptions are, therefore, call (put) options on the
underlying swap rate.
• Swaptions are referred to by their expiry and tenor.
• For example, a 1y2y swaption has a 1 year expiry and the
underlying is a 2 year swap that starts at option expiry.
Note that this is different from caps and floors nomenclature.

MF 728: Fixed Income

Interest Rate Models 26

Swaptions

• Just as we wanted to make the underlying forwards the stochastic

variable in a caplet / Eurodollar option, here the most convenient
approach will be to model the underlying swap.
• It will also be convenient for us to work in a different numeraire
(or measure) which is defined by the annuity function of a swap.
• Recall that the formula for a (par) swap rate is:
PN
\ i=1 δti Lti D(0, ti )
S(t, T) = P N
(14)
i=1 δti D(0, ti )

MF 728: Fixed Income

Interest Rate Models 27

Annuity Numeraire for Swaptions

• The denominator in (14) is the value of a constant stream of
payments and is often referred to as the annuity function:
N
X
A(t, T ) = δti D(t, ti ) (15)
i=1

(here tN = T ).
• The annuity function, or PV01, tells us the present value of a one
basis point annuity between two dates.
• It can be proven that the swap rate is a martingale under the
annuity numeraire.
• Therefore, when we price a swaption, today’s annuity value plays
the role that a discount factor plays in standard risk neutral
valuation for other asset classes.

MF 728: Fixed Income

Interest Rate Models 28

Swaption Pricing in Black / Normal Models

• The price of a payer swaption in the annuity measure can be

written as:
+
 
P = A0 (t, T )E (F − K) (16)

where in this case F refers to the par swap rate.

• To price a swaption, we will need to specify some dynamics for the
underlying swap rate. For example, we could use:
– Black’s Model

dFt = σFt dWt (17)

– Normal Model

dFt = σdWt (18)

MF 728: Fixed Income

Interest Rate Models 29

Swaption Pricing in Black / Normal Models

• If we use either the normal or lognormal model, then we will have

a closed form solution for the price of a payer (or receiver)
swaption in the annuity measure:
– Black’s Model

P = A0 (t, T ) [F0 N (d1 ) − KN (d2 )] (19)

F0

1 2
log K + 2 σ t
d1 = √ (20)
σ t
F0

1 2
log K − 2 σ t
d2 = √ (21)
σ t

MF 728: Fixed Income

Interest Rate Models 30

– Normal Model
√
P = A0 (t, T )σ t [d1 N (d1 ) + N ′ (d1 )] (22)
F0 − K
d1 = √ (23)
σ t
d2 = −d1 (24)

• Note that while these models are simple, they are unlikely to
explain the entire volatility skew for a given underlying swap.
• To incorporate the skew in a robust way, we will need to resort to
a stochastic volatility model, or a jump process.
• Further notice that incorporating skew using a stochastic volatility
or jump process only helps us connect the volatility of different
strikes on the same underlying swap.
• We will need then need to employ other tools to connect the
volatilities of different underlying swaps.

MF 728: Fixed Income

Interest Rate Models 31

Pros and Cons of Normal / Black’s Model

for Swaption Pricing

• The Black and Bachelier models are both simple models with
closed form solutions for their pricing equations.
• Both models rely on a single constant volatility parameter, σ.
• As a result, neither model will enable us to match the volatility
skews that we observe in markets.
• The Bachelier model permits negative rates, whereas Black’s
model does not.
• For swaptions, the market standard quoting convention is to use
the Bachelier model rather than Black’s model, unlike equity
markets.

MF 728: Fixed Income

Interest Rate Models 32

Swaption Pricing: CEV Model

dFt = σFtβ dW (25)

• The CEV model is a generalization of the log-normal and normal

models.
• Model parameters σ and β
• The additional parameter, β can be used to account for some
degree of skew in the volatility surface.
• Cases β = 1 and β = 0 reduce back to the log-normal/normal
models respectively.
• CEV model prices can be obtained via an asymptotic
approximation.

MF 728: Fixed Income

Interest Rate Models 33

Swaption Pricing: Incorporating the

Volatility Smile

• The two most common approaches for fitting the volatility skew in
rates markets are via stochastic volatility models and jump
diffusion models.
• Stochastic Volatility Models: add a second, and potentially
correlated process to the model that controls volatility.
• Jump Diffusion Models: allows the process to jump rather than
follow continuous paths.
• Both of these techniques help to put more mass in the tails of the
distribution, either symmetrically, or asymmetrically.
• The market standard stochastic model in rates is SABR, which, as
we will see next, is a stochastic volatility model.

MF 728: Fixed Income

Interest Rate Models 34

Swaption Pricing: Incorporating the

Volatility Smile

• In rates the standard way of incorporating skew is via the SABR

model, whose dynamics can be written as:

dFt = σt Ftβ dWt1

dσt = ασt dWt2
Cov(dWt1 , dWt2 ) = ρ dt (26)

• Unlike the Black and Normal models, there will be no closed form
solution to the SABR model, however, there are well-documented
implied volatility approximation formulas.
• These formulas give an implied volatility for a given combination
of SABR parameters (α, β, ρ and σ0 ), strike and expiry.

MF 728: Fixed Income

Interest Rate Models 35

SABR Model: Overview

• The SABR model is defined by the following four parameters:
– α: Volatility of volatility
– β: CEV exponent
– ρ: Correlation between interest rates and volatility.
– σ0 : starting level of the volatility process.
• Generally speaking, ρ defines the slope of the volatility skew.
Why?
• α is closely related to the kurtosis in the distribution, and larger α
will lead to fatter tails.
• Unless β is set to zero, SABR does not permit negative rates.
• If we use β = 1 or β = 0 then the forward rate stochastic process
simplifies to Log-Normal / Normal.

MF 728: Fixed Income

Interest Rate Models 36

SABR Model: Asymptotic Approximation

• The following asymptotic formula can be used to compute the

Bachelier or Normal volatility of an option under the SABR model:
F0 − K
σn (T, K, F0 , σ0 , α, β, ρ) = α × ...
∆(K, F0 , σ0 , α, β)
2 2
   
2γ2 − γ1 σ0 C(Fmid ) 2 ργ1 σ0 C(Fmid ) 2 − 3ρ
1+ ( ) + + ϵ (27)
24 α 4 α 24

MF 728: Fixed Income

Interest Rate Models 37

Where:

C(F ) = Fβ
C ′ (Fmid )
γ1 =
C(Fmid )
C ′′ (Fmid )
γ2 =
C(Fmid )
F0 + K
Fmid =
2
p !
1 − 2ρζ + ζ 2 + ζ − ρ
∆(K, F0 , σ0 , α, β) = log
1−ρ
α 
1−β

ζ = F0 − K 1−β
σ0 (1 − β)
ϵ = T α2

• A similar expansion exists for Black’s model.

MF 728: Fixed Income

Interest Rate Models 38

SABR Model: Asymptotic Formula for ATM

Options

• When pricing an at-the-money option, F0 = K, an astute observer

might notice that both the numerator and the denominator are
equal to zero.
• Clearly this is not ideal, as at-the-money options are usually of the
most interest.
• Thankfully, we can apply L’Hopital’s rule to handle these
situations. That is:
0∂F −K
F0 − K
lim α = lim α ∂∆(K,F∂F,σ0
(28)
F0 →K ∆(K, F0 , σ0 , α, β) F0 →K 0 0 ,α,β)
∂F0

MF 728: Fixed Income

Interest Rate Models 39

SABR Model: Calibration

• The SABR model allows for a fairly good fit of the volatility
surface for a single expiry.
• The SABR model relies on separate calibration results for each
expiry.
• When calibrating a SABR model, we need to find the best
parameters (α, β, ρ, σ0 ) for our data.
• Although we have four distinct parameters, there is some
redundancy. This presents challenges in calibration.
• As a result, we often fix one parameter, generally β, and calibrate
the three remaining parameters.

MF 728: Fixed Income

Interest Rate Models 40

SABR Model: Calibration

• A calibration process involves minimizing distance between market

data and model prices.
• For example, we could use the following least squares approach:
 
X 
2
p⃗min = argmin (ĉ(K, p⃗) − cK ) (29)
p
⃗ 
τ,K


where p⃗ = (α, β, ρ, σ0 ); cK and ĉ(K, p⃗) are market and model

option prices respectively.
• In the case of the SABR model, ĉ(K, p⃗) would be obtained using
the SABR approximation formula on the previous slide.

MF 728: Fixed Income

Interest Rate Models 41

SABR Model: An Alternate Approach to

Selecting Beta

• If we consider at-the-money options in Black’s asymptotic SABR

formula, we find the following relationship:

log σAT M ≈ log α − (1 − β) log F (30)

• This provides another way that we can estimate the β parameter

• We can use linear regression on a time series of the log of
at-the-money implied vol against the log of the forward rate.
• Note that we need to apply L’Hopital’s rule in order to use the
asymptotic formula on at-the-money options.

MF 728: Fixed Income

Interest Rate Models 42

SABR Model for Pricing Swaptions

• Recall that the price of a payer swaption in the annuity measure

can be written as:
+
 
P = A0 (t, T )E (S − K) (31)

• Applying the SABR model means that we assume the par swap
rate, S, follows the process in (26).
• To proceed with pricing we need to specify SABR parameters and
apply the SABR approximation formula to obtain a normal /
Bachelier volatility.
• Once we’ve done this, we can obtain a swaption price using (22).

MF 728: Fixed Income

Interest Rate Models 43

Swaption Volatility Cube

• Earlier in this course, we saw that in equity markets the volatility

surface was a two dimensional object (across strike and expiry).
• In rates, we have another dimension that arises from the tenor or
length of a swap. This three dimensional object is referred to as
the volatility cube and has dimensions:
– Expiry
– Tenor
– Strike
• Clearly the volatilities for swaptions with different tenors but the
same strike and expiry should be somewhat related.
• Because of the extra dimension, fitting an interest rate volatility
model is a more complex endeavor than it is for other asset classes.

MF 728: Fixed Income

Interest Rate Models 44

Modeling the Volatility Cube

• In order to model the volatility cube we must come up with a

pricing model that can match prices (or implied volatilities) across
all three dimension
• NOTE: If we are pricing a single swaption or cap, we will not need
to rely on these ”bigger” models for pricing, however, if we are
interested in pricing exotics or computing consistent risk metrics,
then we will need to apply one of these techniques.
• There are two main approaches to modeling the volatility cube:
– Short Rate Models
– Forward Rate (Libor) Market Models

MF 728: Fixed Income

Interest Rate Models 45

Difference between Caps and Swaptions

• Recall that the pricing equation for a cap can be written as:
X
+
 
C(K) = δP (0, Ti + δ)E (FTi − K) (32)
i

Switching sum and expectation we have:

X
+
 
C(K) = E {δP (0, Ti + δ) (FTi − K) } (33)
i

• The pricing equation for a swaption can be written as:

+
 
P = A0 (t, T )E (S − K) (34)
PN
\ i=1 δti Lti D(0, ti )
S(t, T) = P N
(35)
i=1 δti D(0, ti )

• Question: Are these trades the same? What is the difference?

MF 728: Fixed Income

Interest Rate Models 46

Trading Caps vs. Swaptions: The Wedge

Trade

• The difference between a cap and a swaption is akin to the

difference between a basket of options and an option on a basket.
• In other words, the cap in (32) and the swaption in (34) are not
the same, and the gap between them is driven by forward rate
correlations.
• In practice this is a commonly traded structure as it gives investors
a precise way to bet on these correlations and also harvest any
dislocations between the two markets.

MF 728: Fixed Income

Interest Rate Models 47

Mid-Curves

• Mid-curves are options on a forward starting swap.

• Because they are options on a forward starting swap, there is
another variable that defines a mid-curve, namely how far forward
the swap starts.
• Mid-curves are referenced by their expiry / forwardness / tenor
combination.
• For example, a 1y2y3y mid-curve refers to a one year expiry on a
three year swap which starts 2 years after the option expiry.
• Pricing a mid-curve is conceptually equivalent to pricing a
swaption, and is generally done under the annuity measure.

MF 728: Fixed Income

Interest Rate Models 48

Mid-Curves

• Note that the exercise decision on a mid-curve must be made at

expiry and the forwardness and tenor tell us which part of the
curve the option is on.
• Clearly, the volatility between mid-curves and swaptions with the
same or similar expiries must be related.
• Mid-curves provide a convenient instrument for trading forward
volatility.
• In fact, it turns out that a portfolio of swaptions and mid-curves
can be put together to form interesting correlation or dispersion
trades.

MF 728: Fixed Income

Interest Rate Models 49

Correlation Products: Basics of Vanilla vs.

Mid-Curve Triangles

• Using a swap and a forward starting swap we can decompose a

swap into two components.
• In the case of swaps this decomposition is generally not interesting.
• However, this decomposition can also be done with swaptions and
mid-curves.
• In options space this is a way to isolate the correlation or
dispersion of the two pieces vs. the whole. (Why?)

MF 728: Fixed Income

Interest Rate Models 50

Correlation Products: Basics of Vanilla vs.

Mid-Curve Triangles

• For example, consider the following portfolio:

– 1y3y Swaption
– 1y2y Swaption
– 1y2y1y Mid-Curve
• Notice that the first item and the last two items have the same
expiry and span the same underlying swap dates.
• This portfolio is essentially trading a basket of options vs. an
option on a basket. The difference in valuation between these two
quantities will depend on correlation.

MF 728: Fixed Income

Interest Rate Models 51

Correlation Products: Spread Options

• Spread options, or options on the difference between two rates

also are a relatively large piece of the OTC rate derivative market.
• Call spread options are called caps, whereas put spread options are
called floors.
• For example, a common spread option might be a 1y2s30s option.
• In this case, the payout would be based on the 30 year swap rate
less the two year swap rate, and the option expiry would be in 1
year.
• These products present an efficient way for an investor to place a
bet on the steepness of the yield curve.

MF 728: Fixed Income

Interest Rate Models 52

Correlation Products: Spread Options

• As spread options are options on the difference of two random

variables, their valuation will clearly depend on correlation.
• To see the intuition behind pricing spread options, recall that for
two random variables X and Y , we have:

σx−y = σx + σy − 2ρσx σy (36)

• This formula is exact under certain assumptions, but does not

account for the volatility smile (among other things).
• In particular, we can see that the higher the correlation between
the underlying swap rates, the cheaper the spread option.
• Spread options can either be priced via a SABR/LMM or other
volatility cube model, or can be priced using a copula approach.

MF 728: Fixed Income