Interest Rate Modelling and Derivative Pricing

Sebastian Schlenkrich

HU Berlin, Department of Mathematics

WS, 2019/20
 Part II

Yield Curves and Linear Products

p. 49
Outline

Static Yield Curve Modelling and Market Conventions

Multi-Curve Discounted Cash Flow Pricing

Linear Market Instruments

Credit-risky and Collateralized Discounting

p. 50
Outline

Static Yield Curve Modelling and Market Conventions

Multi-Curve Discounted Cash Flow Pricing

Linear Market Instruments

Credit-risky and Collateralized Discounting

p. 51
Outline

Static Yield Curve Modelling and Market Conventions

Yield Curve Representations
Overview Market Conventions for Dates and Schedules
Calendars
Business Day Conventions
Rolling Out a Cash Flow Schedule
Day Count Conventions
Fixed Leg Pricing

p. 52
DCF method requires knowledge of today’s ZCB prices

cash flow stream (or leg)

cash flows V1 V2 ... VN

✻ ✻ ✻ ✻ ✻ ✻
✲
pay times T1 T2 ... TN

◮ Assume t = 0 and deterministic cash flows, then

N
X
V (0) = P(0, Ti ) · Vi .
i=1

How do we get today’s ZCB prices P(0, Ti )?

p. 53
Yield curve is fundamental object for interest rate
modelling
◮ A yield curve (YC) at an observation time t is the function of zero
coupon bonds P(t, ·) : [t, ∞) → R+ for maturities T ≥ t.
◮ YCs are typically represented in terms of interest rates (instead of zero
coupon bond prices).
◮ Discretely compounded zero rate curve zp (t, T ) with frequency p, such
that  −p·(T −t)
zp (t, T )
P(t, T ) = 1+ .
p

◮ Simple compounded zero rate curve z0 (t, T ) (i.e. p = 1/(T − t)), such
that
1
P(t, T ) = .
1 + z0 (t, T ) · (T − t)

◮ Continuous compounded zero rate curve z(t, T ) (i.e. p = ∞), such that
P(t, T ) = exp {−z(t, T ) · (T − t)} .

p. 54
For interest rate modelling we also need continuous
compounded forward rates

Definition (Continuous Forward Rate)

Suppose a given observation time t and zero bond curve P(t, ·) : [t, ∞) → R+
for maturities T ≥ t. The continuous compounded forward rate curve is given
by
∂ ln (P(t, T ))
f (t, T ) = − .
∂T

From the definition follows

 Z T 
P(t, T ) = exp − f (t, s)ds .
t

◮ For static yield curve modelling and (simple) linear instrument pricing we
are interested particularly in curves at t = 0.
◮ For (more complex) option pricing we are interested in modelling curves
at t > 0.

p. 55
We show a typical yield curve example

p. 56
The market data for curve calibration is quoted by market
data providers

p. 57
Outline

Static Yield Curve Modelling and Market Conventions

Yield Curve Representations
Overview Market Conventions for Dates and Schedules
Calendars
Business Day Conventions
Rolling Out a Cash Flow Schedule
Day Count Conventions
Fixed Leg Pricing

p. 58
Recall the introductory swap example

Interbank swap deal example

Dates
Market conventions

How do we get from description to cash flow stream?

p. 59
There are a couple of market conventions that need to be
taken into account in practice

◮ Holiday calendars define at which dates payments can be made.

◮ Business day conventions specify how dates are adjusted if they fall
on a non-business day.

◮ Schedule generation rules specify how regular dates are calculated.

◮ Day count conventions define how time is meassured between dates.

p. 60
Outline

Static Yield Curve Modelling and Market Conventions

Yield Curve Representations
Overview Market Conventions for Dates and Schedules
Calendars
Business Day Conventions
Rolling Out a Cash Flow Schedule
Day Count Conventions
Fixed Leg Pricing

p. 61
Dates are represented as triples day/month/year or as
serial numbers

p. 62
A calender specifies business days and non-business days

Holiday Calendar
A holiday calendar C is a set of dates which are defined as holidays or
non-business days.
◮ A particular date d is a non-business day if d ∈ C.
◮ Holiday calendars are specific to a region, country or market segment.
◮ Need to be specified in the context of financial product.
◮ Typically contain weekends and special days of the year.
◮ May be joined (e.g. for multi-currency products), C¯ = C1 ∪ C2 .
◮ Typical examples are TARGET calendar and LONDON calendar.

t ❞ ❞ ❞ ❞ ❞ t t ❞ ✲

p. 63
Outline

Static Yield Curve Modelling and Market Conventions

Yield Curve Representations
Overview Market Conventions for Dates and Schedules
Calendars
Business Day Conventions
Rolling Out a Cash Flow Schedule
Day Count Conventions
Fixed Leg Pricing

p. 64
A business day convention maps non-business days to
adjacent business days
Business Day Convention (BDC)
◮ A business day convention is a function ωC : D → D which maps a date
d ∈ D to another date d̄.
◮ It is applied in conjunction with a calendar C.
◮ Good business days are unchanged, i.e. ωC (d) = d if d ∈/ C.

Following 
ωC (d) = min d̄ ∈ D\C | d̄ ≥ d
✲
❞✲
Preceding  ❞ ❞
✛
t t
ωC (d) = max d̄ ∈ D\C | d̄ ≤ d

Modified
 Following  
ωCFollowing (d), if Month [d] = Month ωCFollowing (d)
ωC (d) =
ωCPreceeding (d), else

p. 65
Outline

Static Yield Curve Modelling and Market Conventions

Yield Curve Representations
Overview Market Conventions for Dates and Schedules
Calendars
Business Day Conventions
Rolling Out a Cash Flow Schedule
Day Count Conventions
Fixed Leg Pricing

p. 66
Schedules represent sets of regular reference dates

p. 67
Schedule generation follows some rules/conventions as well

1. Consider direction of roll-out: forward or backward (relevant for

front/back stubs).
1.1 Forward, roll-out from start (or effective) date to end (or maturity)
date
1.2 Backward, roll-out from end (or maturity) date to start (or
effective) date

2. Roll out unadjusted dates according to frequency or tenor, e.g. annual

frequency or 3 month tenor

3. If first/last period is broken consider short stub or long stub.

3.1 Short stub is an unregular last period smaller then tenor.
3.2 Long stub is an unregular last period larger then tenor
4. Adjust unadjusted dates according to calendar and BDC.

p. 68
Outline

Static Yield Curve Modelling and Market Conventions

Yield Curve Representations
Overview Market Conventions for Dates and Schedules
Calendars
Business Day Conventions
Rolling Out a Cash Flow Schedule
Day Count Conventions
Fixed Leg Pricing

p. 69
Day count conventions map dates to times or year fractions

Day Count Convention

A day count convention is a function τ : D × D → R which measures a time
period between dates in terms of years.

We give some examples:

Act/365 Fixed Convention

τ (d1 , d2 ) = (d2 − d1 ) /365
◮ Typically used to describe time in financial models.

Act/360 Convention
τ (d1 , d2 ) = (d2 − d1 ) /360
◮ Often used for Libor floating rate payments.
τ (d1 , d2 )
✛ ✲
✲

p. 70
30/360 methods are slightly more involved

General 30/360 Method

◮ Consider two dates d1 and d2 represented as triples of day/month/year,
i.e. d1 = [D1 , M1 , Y1 ] and d2 = [D2 , M2 , Y2 ] with D1/2 ∈ {1, . . . , 31},
M1/2 ∈ {1, . . . , 12} and Y1/2 ∈ {1, 2, . . .}.

◮ Obviously, only valid dates are allowed (no Feb. 30 or similar).

◮ Adjust D1 7→ D̄1 and D2 7→ D̄2 according to specific rules.

◮ Calculate


360 · (Y2 − Y1 ) + 12 · (M2 − M1 ) + D̄2 − D̄1
τ (d1 , d2 ) = .
360

p. 71
Some specific 30/360 rules are given below

30/360 Convention (or 30U/360, Bond Basis)

1. D̄1 = min {D1 , 30}.
2. If D̄1 = 30 then D̄2 = min {D2 , 30} else if D̄2 = D2 .

30E/360 Convention (or Eurobond)

1. D̄1 = min {D1 , 30}.
2. D̄2 = min {D2 , 30}.

p. 72
Outline

Static Yield Curve Modelling and Market Conventions

Yield Curve Representations
Overview Market Conventions for Dates and Schedules
Calendars
Business Day Conventions
Rolling Out a Cash Flow Schedule
Day Count Conventions
Fixed Leg Pricing

p. 73
Now we have all pieces to price a deterministic coupon leg
Coupon is calculated as

Coupon = Notional × Rate × YearFraction

= 100, 000, 000EUR × 3% × τ

p. 74
Outline

Static Yield Curve Modelling and Market Conventions

Multi-Curve Discounted Cash Flow Pricing

Linear Market Instruments

Credit-risky and Collateralized Discounting

p. 75
Outline

Multi-Curve Discounted Cash Flow Pricing

Classical Interbank Floating Rates
Tenor-basis Modelling
Projection Curves and Multi-Curve Pricing

p. 76
Recall the introductory swap example

Stochastic interest rates

How do we model floating rates?

p. 77
We start with some introductory remarks
◮ London Interbank Offered Rates (Libor) currently are the key building
blocks of interest rate derivatives.

◮ They are fixed for USD, GBP, JPY, CHF (and EUR).
◮ EUR equivalent rate is Euribor rate (we will use Libor synonymously for
Euribor).

◮ Libor rate modelling has undergone significant changes since financial

crisis in 2008.

◮ This is typically reflected by the term Multi-Curve Interest Rate

Modelling.

◮ Recent developments in the market will lead to a shift from LIbor rates to
alternative floating rates in the near future (Ibor Transition).

◮ We will also touch on potential new alternative rates specifications when

discussing OIS swaps.

p. 78
Let’s start with the classical Libor rate model

What is the fair interest rate K bank A and Bank B can agree on?

Bank A (lends 1 EUR at T0 ) ✻1 EUR × K × τ

✻1 EUR

T0 τ = τ (T0 , T1 )
✲
Trade agreed at T T1

1 EUR ❄
Bank B (returns 1 EUR plus interest at T1 )
We get (via DCF methodology)

0 = V (T ) = P(T , T0 ) · ET0 [−1 | FT ] + P(T , T1 ) · ET1 [1 + τ K | FT ] ,

0 = −P(T , T0 ) + P(T , T1 ) · (1 + τ K ) .

p. 79
Spot Libor rates are fixed daily and quoted in the market

0 = −P(T , T0 ) + P(T , T1 ) · (1 + τ K )

Spot Libor rate

The fair rate for an interbank lending deal with trade date T , spot starting
date T0 (typically 0d or 2d after T ) and maturity date T1 is
 
P(T , T0 ) 1
L(T ; T0 , T1 ) = −1 .
P(T , T1 ) τ

◮ Panel banks submit daily estimates for interbank lending rates to

calculation agent.

◮ Relevant periods (i.e. [T0 , T1 ]) considered are 1m, 3m, 6m and 12m.
◮ Trimmed average of submissions is calculated and published.
Libor rate fixings currently are the most important reference rates for interest
rate derivatives.

p. 80
Example publication at Intercontinental Exchange (ICE)

p. 81
A plain vanilla Libor leg pays periodic Libor rate coupons

L(T0F ; T0 , T1 ) L(T1F ; T1 , T2 ) F
L(TN−1 ; TN−1 , TN )
✻ ✻ ✻
T0F T1F F
TN−1
... ✲
✛ ✲✛ ✲ ✛ ✲
τ1 τ2 τN
T0 T1 T2 TN−1 TN
We get (via DCF methodology)
N
X  
V (t) = P(t, Ti ) · ETi L(Ti−1
F
; Ti−1 , Ti ) · τi | Ft
i=1
N
X  
= P(t, Ti ) · ETi L(Ti−1
F
; Ti−1 , Ti ) | Ft · τi .
i=1

Thus all we need is  

ETi L(Ti−1
F
; Ti−1 , Ti ) | Ft =?

p. 82
Libor rate is a martingale in the terminal measure
Theorem (Martingale property of Libor rate)
The Libor rate L(T ; T0 , T1 ) with observation/fixing date T , accrual start date
T0 and accrual end date T1 is a martingale in the T1 -forward measure and
 
T1 P(t, T0 ) 1
E [L(T ; T0 , T1 ) | Ft ] = −1 = L(t; T0 , T1 ).
P(t, T1 ) τ

Proof.
The fair Libor rate at fixing time T is L(T ; T0 , T1 ) = [P(T , T0 )/P(T , T1 ) − 1] /τ .
The zero coupon bond P(T , T0 ) is an asset and P(T , T1 ) is the numeraire in the
T1 -forward meassure. Thus FTAP yields that the discounted asset price is a
martingale, i.e. h i
P(T , T0 ) P(t, T0 )
ET1 | Ft = ,
P(T , T1 ) P(t, T1 )
Linearity of expectation operator yields
h h i i h i
P(T , T0 ) 1 P(t, T0 ) 1
ET1 [L(T ; T0 , T1 ) | Ft ] = ET1 |Ft − 1 = −1 = L(t; T0 , T1 ).
P(T , T1 ) τ P(t, T1 ) τ

p. 83
This allows pricing the Libor leg based on today’s
knowledge of the yield curve only

L(T0F ; T0 , T1 ) L(T1F ; T1 , T2 ) F
L(TN−1 ; TN−1 , TN )
✻ ✻ ✻
T0F T1F F
TN−1
... ✲
✛ ✲✛ ✲ ✛ ✲
τ1 τ2 τN
T0 T1 T2 TN−1 TN
Libor leg becomes
N
X  
V (t) = P(t, Ti ) · ETi L(Ti−1
F
; Ti−1 , Ti ) · τi | Ft
i=1
N
X
= P(t, Ti ) · L(t; Ti−1 , Ti ) · τi
i=1

p. 84
Libor leg may be simplified in the current single-curve
setting
We have
N
X
V (t) = P(t, Ti ) · L(t; Ti−1 , Ti ) · τi
i=1
with  
P(t, Ti−1 ) 1
L(t; Ti−1 , Ti ) = −1 .
P(t, Ti ) τi
This yields
N  
X P(t, Ti−1 ) 1
V (t) = P(t, Ti ) · −1 · τi
P(t, Ti ) τi
i=1
N
X
= P(t, Ti−1 ) − P(t, Ti )
i=1

= P(t, T0 ) − P(t, TN ).

We only need discount fators P(t, T0 ) and P(t, TN ) at first date T0 and last
date TN .
p. 85
Outline

Multi-Curve Discounted Cash Flow Pricing

Classical Interbank Floating Rates
Tenor-basis Modelling
Projection Curves and Multi-Curve Pricing

p. 86
The classical Libor rate model misses an important detail

What if a counterparty defaults?

p. 87
What if Bank B defaults prior to T0 or T1 ?
What is the fair rate K bank A and Bank B can agree on
given the risk of default?

Bank A (lends 1 EUR at T0 ) ✶{ξB >T1 } ✻

× 1 EUR × K × τ
✶{ξB >T1 } ✻
× 1 EUR

T0 τ = τ (T0 , T1 )
✲
Trade agreed at T T1

✶{ξB >T0 } × 1 EUR ❄

Bank B (returns 1 EUR plus interest at T1 )

◮ Cash flows are paid only if no default occurs.

◮ We apply a simple credit model.
◮ Denote ✶D the indicator function for an event D and random variable ξB
the first time bank B defaults.

p. 88
Credit-risky trade value can be derived using derivative
pricing formula
 
V (T ) 1 1+K ·τ
= EQ −✶{ξB >T0 } · + ✶{ξB >T1 } · .
B(T ) B(T0 ) B(T1 )
(all expectations conditional on FT )

Assume independence of credit event ξB > T0/1 and interest rate market,
then
   
V (T )   1   1+K ·τ
= −EQ ✶{ξB >T0 } · EQ + EQ ✶{ξB >T1 } · EQ .
B(T ) B(T0 ) B(T1 )

h i
Abbreviate survival probability Q(T , T0,1 ) = EQ ✶{ξB >T0,1 } | FT and apply
change of measure

V (T ) = −P(T , T0 )Q(T , T0 )ET0 [1] + P(T , T1 )Q(T , T1 )ET1 [1 + K · τ ] .

p. 89
This yields the fair spot rate in the presence of credit risk

V (T ) = −P(T , T0 )Q(T , T0 )ET0 [1] + P(T , T1 )Q(T , T1 )ET1 [1 + K · τ ] .

If we solve V (T ) = 0 and set K = L(T ; T0 , T1 ) we get

 
P(T , T0 ) Q(T , T0 ) 1
L(T ; T0 , T1 ) = · −1 .
P(T , T1 ) Q(T , T1 ) τ

We need a model for the survival probability Q(T , T1,2 ).

n R o
T1,2
Consider, e.g., hazard rate model Q(T , T1,2 ) = exp − T
λ(s)ds with
deterministic hazard rate λ(s). Then forward survival probability D(T0 , T1 )
with
 Z T1 
Q(T , T0 )
D(T0 , T1 ) = = exp − λ(s)ds
Q(T , T1 ) T0

is independent of observation time T .

p. 90
Deterministic hazard rate assumption preserves the
martingale property of forward Libor rate

Theorem (Martingale property of credit-risky Libor rate)

Consider the credit-risky Libor rate L(T ; T0 , T1 ) with observation/fixing date
T , accrual start date T0 and accrual end date T1 . If the forward survival
probability D(T0 , T1 ) is deterministic such that
 
P(T , T0 ) 1
L(T ; T0 , T1 ) = · D(T0 , T1 ) − 1 ,
P(T , T1 ) τ

then L(t; T0 , T1 ) is a martingale in the T1 -forward measure and

 
P(t, T0 ) 1
ET1 [L(T ; T0 , T1 ) | Ft ] = L(t; T0 , T1 ) = · D(T0 , T1 ) − 1 .
P(t, T1 ) τ

Proof.
Follows analogously to classical Libor rate martingale property.

p. 91
Outline

Multi-Curve Discounted Cash Flow Pricing

Classical Interbank Floating Rates
Tenor-basis Modelling
Projection Curves and Multi-Curve Pricing

p. 92
Forward Libor rates are typically parametrised via
projection curve n R o
T1,2
◮ Hazard rate λ(u) in Q(T , T1,2 ) = exp − T
λ(u)du is often
considered as a tenor basis spread s(u).
◮ Survival probability Q(T , T1,2 ) can be interpreted as discount factor.
◮ Suppose we know time-t survival probabilities Q(t, ·) for a forward Libor
rate L(t, T0 , T0 + δ) with tenor δ (typically 1m, 3m, 6m or 12m). Then
we define the projection curve

P δ (t, T ) = P(t, T ) · Q(t, T ).

◮ With projection curve P δ (t, T ) the forward Libor rate formula is

analogous to the classical Libor rate formula, i.e.
 
P δ (t, T0 ) 1
Lδ (t, T0 ) = L(t; T0 , T0 + δ) = −1 .
P δ (t, T1 ) τ

This yields the multi-curve modelling framework consisting of discount curve

P(t, T ) and tenor-dependent projection curves P δ (t, T ).
p. 93
There is an alternative approach to multi-curve modelling

Define forward Libor rate Lδ (t, T0 ) for a tenor δ as

Lδ (t, T0 ) = ET1 [L(T ; T0 , T0 + δ) | Ft ] .

(Without any assumptions on default, survival probabilities etc.)

Postulate a projection curve parametrisation
 
P δ (t, T0 ) 1
Lδ (t, T0 ) = −1 .
P δ (t, T1 ) τ

◮ We will discuss calibration of projection curve P δ (t, T ) later.

◮ This approach alone suffices for linear products (e.g. Libor legs) and
simple options.

◮ It does not specify any relation between projection curve P δ (t, T ) and
discount curve P(t, T ).

p. 94
Projection curves can also be written in terms of zero rates
and continuous forward rates

Consider a projection curve given by (pseudo) discount factors P δ (t, T )

(observed today).
◮ Corresponding continuous compounded zero rates are
 
δ ln P δ (t, T )
z (t, T ) = − .
T −t

◮ Corresponding continuous compounded forward rates are

 
δ ∂ ln P δ (t, T )
f (t, T ) = − .
∂T

p. 95
We illustrate an example of a multi-curve set-up for EUR

Market data as of July 2016

p. 96
Libor leg pricing needs to be adapted slightly for
multi-curve pricing
Classical single-curve Libor leg price is
N
X
V (t) = P(t, Ti ) · L(t; Ti−1 , Ti ) · τi
i=1

= P(t, T0 ) − P(t, TN ).

Multi-curve Libor leg pricing becomes

N
X
V (t) = P(t, Ti ) · Lδ (t, Ti−1 ) · τi
i=1

with  
P δ (t, Ti−1 ) 1
Lδ (t, Ti−1 ) = −1 .
P δ (t, Ti ) τi
◮ Note that we need different yield curves for Libor rate projection and cash
flow discounting.
◮ Single-curve pricing formula simplification does not work for multi-curve
pricing.
p. 97
Outline

Static Yield Curve Modelling and Market Conventions

Multi-Curve Discounted Cash Flow Pricing

Linear Market Instruments

Credit-risky and Collateralized Discounting

p. 98
Outline

Linear Market Instruments

Vanilla Interest Rate Swap
Forward Rate Agreement (FRA)
Overnight Index Swap
Summary linear products pricing

p. 99
With the fixed leg and Libor leg pricing available we can
directly price a Vanilla interest rate swap

float leg (EUR conventions: 6m Euribor, Act/360)

L1 Lm
✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻
T̃0
✛τ̃j✲ T̃m ✲
T0 ✛ ✲ Tn
τi
❄ ❄ ❄ ❄ ❄ ❄
K K

fixed leg (EUR conventions: annual, 30/360)

Present value of (fixed rate) payer swap with notional N becomes

m n
X X
V (t) = N · L6m (t, T̃j−1 ) · τ̃j · P(t, T̃j ) − N · K · τi · P(t, Ti ).
j=1 i=1

p. 100
Vanilla swap pricing formula allows us to price the
underlying swap of our introductory example
Interbank swap deal example

p. 101
We illustrate swap pricing with QuantLib/Excel...

p. 102
Outline

Linear Market Instruments

Vanilla Interest Rate Swap
Forward Rate Agreement (FRA)
Overnight Index Swap
Summary linear products pricing

p. 103
Forward Rate Agreement yields exposure to single forward
Libor rates
✛ ✲
floating rate τ Lδ (TF ) τ
1+τ Lδ (TF ) ✻ ✻
✙ ✲
t TF ❨
τK
T0 T0 + δ
fixed rate payment 1+τ Lδ (TF )
❄ ❄

◮ Fixed rate K agreed at trade inception (prior to t).

◮ Libor rate Lδ (TF , T0 ) fixed at TF , valid for the period T0 to T0 + δ.
 
◮ Payoff paid at T0 is difference τ · Lδ (TF , T0 ) − K discounted from T1
 −1
to T0 with discount factor 1 + τ · Lδ (TF , T0 ) , i.e.
 
τ · Lδ (TF , T0 ) − K
V (T0 ) = .
1 + τ · Lδ (TF , T0 )

p. 104
Time-TF FRA price can be obtained via deterministic basis
spread model
τ ·[Lδ (TF ,T0 )−K ]
Note that payoff V (T0 ) = 1+τ ·Lδ (TF ,T0 )
is already determined at TF .
Thus (via DCF)
 
τ · Lδ (TF , T0 ) − K
V (TF ) = P(TF , T0 ) · V (T0 ) = P(TF , T0 ) · .
1 + τ · Lδ (TF , T0 )

Recall that (with T1 = T0 + δ)

P δ (TF , T0 ) P(TF , T0 )
1 + τ · Lδ (TF , T0 ) = = · D(T0 , T1 ).
P δ (TF , T1 ) P(TF , T1 )

Then
 1
 P(TF , T1 )
V (TF ) = P(TF , T0 ) · τ · Lδ (TF , T0 ) − K · ·
D(T0 , T1 ) P(TF , T0 )
  1
= P(TF , T1 ) · τ · Lδ (TF , T0 ) − K · .
D(T0 , T1 )

p. 105
Present value of FRA can be obtained via martingale
property
Derivative pricing formula in T1 -terminal measure yields
 
V (t) P(TF , T1 )   1
= ET1 · τ · Lδ (TF , T0 ) − K ·
P(t, T1 ) P(TF , T1 ) D(T0 , T1 )
    1
= τ · ET1 Lδ (TF , T0 ) − K ·
D(T0 , T1 )
  1
= τ · Lδ (t, T0 ) − K · .
D(T0 , T1 )
P(t,T0 )
Using 1 + τ · Lδ (t, T0 ) = P(t,T1 )
· D(T0 , T1 ) (deterministic spread assumption)
yields
 −1
 δ P(t, T0 ) 
V (t) = P(t, T0 ) · τ · L (t, T0 ) − K · · D(T0 , T1 )
P(t, T1 )
 
Lδ (t, T0 ) − K · τ
= P(t, T0 ) · .
1 + τ · Lδ (t, T0 )

p. 106
Outline

Linear Market Instruments

Vanilla Interest Rate Swap
Forward Rate Agreement (FRA)
Overnight Index Swap
Summary linear products pricing

p. 107
Overnight index swap (OIS) instruments are further
relevant instruments in the market
compounding leg
C1 ✻ ...✻ Cm ✻

✲
T0 T1 ...
fixed leg accrual dates T0 , T1
K ❄ K ❄ K ❄

compounding leg coupon with compounding rate C1 Qk 

i=1
(1+Li τi ) −1 ✻
observation dates t0 , . . . , tk C1 = τ (T0 ,T1 )

overnight rates Li = L(ti−1 ; ti−1 , ti )

✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻
❘ ❘ ❘ ❘ ❘ ❘ ❘ ❘ ❘ ❘ ✲
t0 = T0 t1 t2 ...
✛✲ tk−1 tk = T1
τi = 1d

p. 108
We need to calculate the compounding leg coupon rate
◮ Assume overnight index swap (OIS) rate Li = L(ti−1 ; ti−1 , ti ) is a
credit-risk free Libor rate.

◮ Compounded rate (for a period [T0 , T1 ]) is specified as

(" k # )
Y 1
C1 = (1 + Li τi ) − 1 .
τ (T0 , T1 )
i=1

◮ Coupon payment is at T1 .
◮ For pricing we need to calculate
"(" k # ) #
T1 T1
Y 1
E [C1 | Ft ] = E (1 + Li τi ) − 1 | Ft
τ (T0 , T1 )
i=1
( " k
# )
T1
Y 1
= E (1 + Li τi ) | Ft − 1 .
τ (T0 , T1 )
i=1

p. 109
How do we handle the compounding term?
Overall compounding term is
k k
Y Y
(1 + Li τi ) = [1 + L(ti−1 ; ti−1 , ti )τi ] .
i=1 i=1

Individual compounding term is

 
P(ti−1 , ti−1 ) 1 P(ti−1 , ti−1 )
1 + L(ti−1 ; ti−1 , ti )τi = 1 + −1 τi = .
P(ti−1 , ti ) τi P(ti−1 , ti )

We get
k k k
Y Y P(ti−1 , ti−1 ) Y 1
(1 + Li τi ) = = .
P(ti−1 , ti ) P(ti−1 , ti )
i=1 i=1 i=1

Qk 1
We need to calculate the expectation of i=1 P(ti−1 ,ti )
.

p. 110
Expected compounding factor can be easily calculated

Lemma (Compounding rate)

Consider a compounding coupon period [T0 , T1 ] with overnight observation
and maturity dates {t0 , t1 , . . . , tk }, t0 = T0 and tk = T1 . Then
" k
#
T1
Y 1 1
E | FT0 = .
P(ti−1 , ti ) P(T0 , T1 )
i=1

p. 111
We proof the result via Tower Law of conditional
expectation

" k
# " " k
# #
Y 1 Y 1
T1 T1 T1
E | FT0 = E E | Ftk−2 | FT0
P(ti−1 , ti ) P(ti−1 , ti )
i=1 i=1
"k−1   #
Y 1 P(tk−1 , tk−1 )
T1 T1
= E E | Ftk−2 | FT0
P(ti−1 , ti ) P(tk−1 , tk )
i=1
"k−1 #
Y 1 P(tk−2 , tk−1 )
T1
= E | FT0
P(ti−1 , ti ) P(tk−2 , tk )
i=1
"k−2 #
Y 1 1
T1
= E | FT0
P(ti−1 , ti ) P(tk−2 , tk )
i=1
h i
1
... = ET1 | FT0
P(t0 , tk )
1
= .
P(T0 , T1 )

p. 112
Expected compounding rate equals Libor rate

◮ Expected compounding rate as seen at start date T0 becomes

 
T1 1 1
E [C1 | FT0 ] = −1 = L(T0 ; T0 , T1 ).
P(T0 , T1 ) τ (T0 , T1 )

◮ Consequently, expected compounding rate equals Libor rate for full

period.

◮ Moreover, expectations as seen of time-t are

" k
#
T1
Y 1 P(t, T0 )
E | Ft =
P(ti−1 , ti ) P(t, T1 )
i=1

and  
T1 P(t, T0 ) 1
E [C1 | Ft ] = −1 = L(t; T0 , T1 ).
P(t, T1 ) τ (T0 , T1 )

p. 113
Compounding swap pricing is analogous to Vanilla swap
pricing

compounding leg
C1 ✻ ...✻ Cm ✻

✲
T0 T1 ...
fixed leg
K ❄ K ❄ K ❄

m
X m
X
V (t) = N · ETj [Cj | Ft ] · τj · P(t, Tj ) − N · K · τj · P(t, Tj )
j=1 j=1
m m
X X
= N · L(t; Tj−1 , Tj ) · τj · P(t, Tj ) − N · K · τj · P(t, Tj ).
j=1 j=1

p. 114
Outline

Linear Market Instruments

Vanilla Interest Rate Swap
Forward Rate Agreement (FRA)
Overnight Index Swap
Summary linear products pricing

p. 115
As a summary we give an overview of linear products
pricing
Vanilla (Payer) Swap
m n
X X
Swap(t) = N · Lδ (t, T̃j−1 ) · τ̃j · P(t, T̃j ) − N · K · τi · P(t, Ti )
j=1 i=1
| {z } | {z }
float leg fixed Leg

Market Forward Rate Agreement (FRA)

  1
FRA(t) = P(t, T0 ) · Lδ (t, T0 ) − K · τ ·
| {z } | {z } 1 + τ · Lδ (t, T0 )
discounting to T0
| {z }
payoff
discounting from T0 to T0 +δ

Compounding Swap / OIS Swap

m m
X X
CompSwap(t) = N · L(t; Tj−1 , Tj ) · τj · P(t, Tj ) − N · K · τj · P(t, Tj )
j=1 j=1
| {z } | {z }
compounding leg fixed leg

p. 116
Further reading on yield curves, conventions and linear
products

◮ F. Ametrano and M. Bianchetti. Everything you always wanted to know

about Multiple Interest Rate Curve Bootstrapping but were afraid to ask
(April 2, 2013).
Available at SSRN: http://ssrn.com/abstract=2219548 or
http://dx.doi.org/10.2139/ssrn.2219548, 2013

◮ M. Henrard. Interest rate instruments and market conventions guide 2.0.

Open Gamma Quantitative Research, 2013

◮ P. Hagan and G. West. Interpolation methods for curve construction.

Applied Mathematical Finance, 13(2):89–128, 2006

On current discussion of Libor alternatives, e.g.

◮ M. Henrard. A quant perspective on ibor fallback proposals.

https://ssrn.com/abstract=3226183, 2018

p. 117
Outline

Static Yield Curve Modelling and Market Conventions

Multi-Curve Discounted Cash Flow Pricing

Linear Market Instruments

Credit-risky and Collateralized Discounting

p. 118
So far we discussed risk-free discount curves and tenor
forward curves - now it is getting a bit more complex

6m projection curve for

L6m (t, T )

✻
Risk-free curve for 3m projection curve for
✲
P(t, T ) L3m (t, T )
❳❳❳
❅ ❳❳❳
❅ ❳❳❳
❘
❅ ❳❳
③
Credit-risky discount Collateral discount
curves curves

Specifying appropriate discount and projection curves for a financial

instrument is an important task in practice.

p. 119
Outline

Credit-risky and Collateralized Discounting

Credit-risky Discounting
Collateralized Discounting

p. 120
Discounting of bond or loan cash flows is subject to credit
risk

✶{ξB >TN } · (1 + K τ )
Investor lends 1 EUR notional ✻
to bank at T0 ✶{ξB >Ti } · K τ ✻

T0 ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻ ✻
✲
T1 T2 ... TN
Bank returns perodic interest K · τ at T1 , . . . , TN
❄ and 1 EUR notional at TN

◮ Cash flows are paid only if no default occurs.

◮ Denote ✶D the indicator function for an event D and random variable ξB
the first time bank defaults.
◮ Assume independence of credit event {ξB > T } and interest rate market

p. 121
We repeat credit-risky valuation from multi-curve pricing
Consider an observation time t with T0 < t ≤ TN then present value of bond
cash flows becomes
" #
V (t) 1 X Kτ
= EQ ✶{ξB >TN } + ✶{ξB >Ti } | Ft .
B(t) B(TN ) B(Ti )
Ti ≥t

Independence of credit event {ξB > T } and interest rate market yields (all
expectations conditional on Ft )

   
V (t)   1 X Q  Kτ
= EQ ✶{ξB >TN } EQ + E ✶{ξB >Ti } EQ .
B(t) B(TN ) B(Ti )
Ti ≥t
 
Denote survival probability Q(t, T ) = E Q
✶{ξB >T } | Ft and change to forward
measure, then
X
V (t) = Q(t, TN )P(t, TN ) + Q(t, TN )P(t, TN )K τ.
Ti ≥t

p. 122
Survival probabilities are parameterized in terms of spread
curves - this leads to credit-risky discount curves

Assume survival probability Q(t, T ) is given in terms of a credit spread curve

s(t) and  Z T 
Q(t, T ) = exp − s(u)du .
t
Also recall that discount factors may be represented in terms of forward rates
f (t, T ) and  
Z T
P(t, T ) = exp − f (t, u)du .
t

We may define a credit-risky discount curve P B (t, T ) for a bond or loan as

 Z T 
P B (t, T ) = Q(t, T )P(t, T ) = exp − [f (t, u) + s(u)] du .
t

p. 123
We can adapt the discounted cash flow pricing method to
cash flows subject to credit risk

Present value of bond or loan cash flows become

X
V (t) = P B (t, TN ) + P B (t, TN )K τ.
Ti ≥t

◮ Bonds are issued by many market participants (banks, corporates,

governments, ...)
◮ Credit spread curves and credit-risky discount curves are specific to an
issuer, e.g. Deutsche Bank has a different credit spread than
Bundesrepublik Deutschland
◮ Many bonds are actively traded in the market. Then we may use market
prices and infer credit spreads s(t) and credit-risky discount curves
P B (t, T )

p. 124
Outline

Credit-risky and Collateralized Discounting

Credit-risky Discounting
Collateralized Discounting

p. 125
For derivative transactions credit risk is typically mitigated
by posting collateral
V (0)
✲
✛ ✛
C (0) C (0)

✛ dC (t)✲ ✛ dC (t)
✲
Bank A Bank B
r (t)C (t)dt rC (t)C (t)dt
✲ ✲

C (T )✲ C (T ) ✲
✛
V (T )
Pricing needs to take into account interest payments on collateral.2

2
Collateral amounts C (t) and collateral rates are agreed in Credit Support
Annexes (CSAs) between counterparties. p. 126
Collateralized derivative pricing takes into account
collateral cash flows

Collateralized derivative price is given by (expectation of) sum of discounted

payoff R T
− r (u)du
e t V (T )
plus sum discounted collateral interest payments
Z T Rs
− r (u)du
e t [r (s) − rC (s)] C (s)ds.
t

That gives

 RT Z T Rs 
− r (u)du − r (u)du
V (t) = EQ e t V (T ) + e t [r (s) − rC (s)] C (s)ds | Ft .
t

p. 127
Pricing is reformulated to focus on collateral rate

Theorem (Collateralized Discounting)

Consider the price of an option V (t) at time t which pays an amount V (T ) at
time T ≥ t (and no intermediate cash flows).
The option is assumed collateralized with cash amounts C (s) (for t ≤ s ≤ T ).
For the cash collateral a collateral rate rC (s) (for t ≤ s ≤ T ) is applied.
Then the option price V (t) becomes
 RT 
− rC (u)du
V (t) = EQ e t V (T ) | Ft
Z T Rs 
− rC (u)du
− EQ e t [r (s) − rC (s)] [V (s) − C (s)] ds | Ft
t

For further details on collateralized discounting see, e.g.

◮ V. Piterbarg. Funding beyond discounting: collateral agreements and derivatives pricing.

Asia Risk, pages 97–102, February 2010

◮ M. Fujii, Y. Shimada, and A. Takahashi. Collateral posting and choice of collateral currency - implications
for derivative pricing and risk management (may 8, 2010).
Available at SSRN: https://ssrn.com/abstract=1601866, May 2010

p. 128
Collateralized discounting result is proved in three steps

1. Define the discounted collateralized price process

Rt Z t Rs
− r (u)du − r (u)du
X (t) = e 0 V (t) + e 0 [r (s) − rC (s)] C (s)ds
0

and show that it is a martingale

2. Analyse the dynamics dX (t) and deduce the dynamics for dV (t)

3. Solve the SDE for dV (t) and calculate price via conditional expectation

p. 129
Step 1 - discounted collateralized price process (1/2)

Consider T ≥ t, then
RT Z T Rs
− r (u)du − r (u)du
X (T ) = e 0 V (T ) + e 0 [r (s) − rC (s)] C (s)ds
RT Z0 t Rs
− r (u)du − r (u)du
=e 0 V (T ) + e 0 [r (s) − rC (s)] C (s)ds+
0
Z T Rs
− r (u)du
e 0 [r (s) − rC (s)] C (s)ds
t
Rt  RT Z T Rs 
− r (u)du − r (u)du − r (u)du
=e 0 e t V (T ) + e t [r (s) − rC (s)] C (s)ds +
t
| {z }
K (t,T )
Z t Rs
− r (u)du
e 0 [r (s) − rC (s)] C (s)ds.
0

p. 130
Step 1 - discounted collateralized price process (2/2)
We have from collateralized derivative pricing that
 RT Z T Rs 
Q Q − r (u)du − r (u)du
E [K (t, T ) | Ft ] = E e t V (T ) + e t [r (s) − rC (s)] C (s)ds | Ft
t
= V (t).

This yields
 Rt Z t Rs 
Q Q − r (u)du − r (u)du
E [X (T ) | Ft ] = E e 0 K (t, T ) + e 0 [r (s) − rC (s)] C (s)ds | Ft
0
Rt Z t Rs
− r (u)du − r (u)du
=e 0 EQ [K (t, T ) | Ft ] + e 0 [r (s) − rC (s)] C (s)ds
0
Rt Z t Rs
− r (u)du − r (u)du
=e 0 V (t) + e 0 [r (s) − rC (s)] C (s)ds
0
= X (t).

Thus, X (t) is indeed a martingale.

p. 131
Step 2 - dynamics dX (t) and dV (t)

Rt Rt Rs
− r (u)du − r (u)du
From the definition X (t) = e 0 V (t) + e 0 [r (s) − rC (s)] C (s)ds
0
follows
Rt Rt Rt
− r (u)du − r (u)du − r (u)du
dX (t) = −r (t)e 0 V (t)dt + e 0 dV (t) + e 0 [r (t) − rC (t)] C (t)dt
Rt
− r (u)du
=e 0 [dV (t) − r (t)V (t)dt + [r (t) − rC (t)] C (t)dt]
Rt
− r (u)du
=e 0 [dV (t) − rC (t)V (t)dt + [r (t) − rC (t)] [C (t) − V (t)] dt] .
| {z }
dM(t)

Since X (t) is a martingale we must have that dM(t) are increments of a martingale.
We get

dV (t) = rC (t)V (t)dt − [r (t) − rC (t)] [C (t) − V (t)] dt + dM(t).

p. 132
Step 3 - solution for V (t) (1/2)
For the SDE dV (t) = rC (t)V (t)dt − [r (t) − rC (t)] [C (t) − V (t)] dt + dM(t) we may
guess a solution as
Rt Z t Rt
rC (s)ds rC (u)du
V (t) = e t0 V (t0 ) − e s {[r (s) − rC (s)] [C (s) − V (s)] ds − dM(s)}
t0

Differentiating confirms that

Rt
rC (s)ds
dV (t) = rC (t)e t0 V (t0 )
Z t Rt
rC (u)du
− rC (t) e s {[r (s) − rC (s)] [C (s) − V (s)] ds − dM(s)}
t0
Rt
rC (u)du
−e t {[r (t) − rC (t)] [C (t) − V (t)] dt − dM(t)}
 Rt Z t Rt 
rC (s)ds rC (u)du
= rC (t) e t0 V (t0 ) − e s {[r (s) − rC (s)] [C (s) − V (s)] ds − dM(s)}
t0
− [r (t) − rC (t)] [C (t) − V (t)] dt + dM(t)
= rC (t)V (t) − [r (t) − rC (t)] [C (t) − V (t)] dt + dM(t).

p. 133
Step 3 - solution for V (t) (2/2)
Substituting t 7→ T and t0 7→ t yields the representation
RT Z T RT
rC (s)ds rC (u)du
V (T ) = e t V (t) − e s {[r (s) − rC (s)] [C (s) − V (s)] ds − dM(s)}
t

Solving for V (t) gives

RT Z T Rs
− rC (s)ds − rC (u)du
V (t) = e t V (T ) − e t {[r (s) − rC (s)] [C (s) − V (s)] ds − dM(s)}
t

The result follows now from taking conditional expectation

 RT Z T Rs 
Q − rC (s)ds − rC (u)du
V (t) = E e t V (T ) − e t [r (s) − rC (s)] [V (s) − C (s)] ds | Ft
t
Z T Rs 
− rC (u)du
+ EQ e t dM(s) | Ft
t
| {z }
0

p. 134
A very important special case arises for full collateralization

Corollary (Full collateralization)

If the collateral amount C (s) equals the full option price V (s) for t ≤ s ≤ T
then the derivative price becomes
 RT 
− rC (s)ds
V (t) = EQ e t V (T ) | Ft .

◮ Fully collateralized price is calculated analogous to uncollateralized price.

◮ Discount rate must equal to the collateral rate rC (s).
◮ Pricing is independent of the risk-free rate r (t).
R t
◮ Collateral bank account B C (t) = exp 0
rC (s)ds can be considered as
numeraire in this setting

p. 135
The collateralized zero coupon bond can be used to adapt
DCF method to collateralized derivative pricing
Consider a fully collateralized instrument that pays V (T ) = 1 at some time
horizon T . The price V (t) fort ≤ T is given by
RT
− rC (s)ds
V (t) = EQ e t 1 | Ft .

Definition (Collateralized zero coupon bond)

The collateralized zero coupon bond price (or collateralized discount factor) for
an observation time t and maturity T ≥ t is given by
 RT 
C Q − rC (s)ds
P (t, T ) = E e t | Ft .

Consider a time horizon T and the time-t price process of a collateralized zero
coupon bond P C (t, T ):
◮ Collateralized zero coupon bond is an asset in our economy,
◮ price process P C (t, T ) > 0.
Thus collateralized zero coupon bond is a numeraire.

p. 136
The collateralized zero coupon bond can be used as
numeraire for pricing
Define the collateralized forward measure QT ,C as the equivalent martingale
measure with P C (t, T ) as numeraire and expectation ET ,C [·].
The density process of QT ,C (relative to risk-neutral measure Q) is

P C (t, T ) B C (0)
ζ(t) = · .
B C (t) P C (0, T )
This yields
   
T ,C Q ζ(T ) P C (T , T ) B C (t)
E [V (T ) | Ft ] = E V (T ) | Ft = EQ · V (T ) | Ft
ζ(t) B C (T ) P C (t, T )
 
1 B C (t)
= C EQ · V (T ) | Ft
P (t, T ) B C (T )
 RT 
1 − rC (s)ds
= C EQ e t V (T ) | Ft
P (t, T )
V (t)
=
P C (t, T )

p. 137
Discounted cash flow method pricing requires to use the
appropriate discount curve representing collateral rates

We have
V (t) = P C (t, T ) · ET ,C [V (T ) | Ft ] .
 RT 
◮ Requires discounting curve P C (t, T ) = EQ e − t
rC (s)ds
| Ft capturing
collateral costs and
◮ calculation of expected future payoffs ET ,C [V (T ) | Ft ] in the
collateralized forward measure.

p. 138
We summarise the multi-curve framework widely adopted
in the market
6m projection curve for
L6m (t, T )

✻
Risk-free curve for 3m projection curve for
✲
P(t, T ) L3m (t, T )
❳❳
❅ ❳❳❳
❅ ❳❳❳
❘
❅ ❳❳❳
③
Credit-risky discount Collateral discount
curves P B (t, T ) curves P C (t, T )

◮ Standard collateral curve is typically also considered as the risk-free curve.

◮ Currently standard collateral curves move from Eonia to €STR collateral
rate (EUR) and Fed Fund to SOFR collateral rate (USD).
◮ Projection curves are potentially not required anymore in the future if
Libor and Euribor indices are decommissioned.
p. 139
Contact

Dr. Sebastian Schlenkrich

Office: RUD25, R 1.211
Mail: sebastian.schlenkrich@hu-berlin.de

d-fine GmbH
Mobile: +49-162-263-1525
Mail: sebastian.schlenkrich@d-fine.de