Professional Documents
Culture Documents
Chapter 1 - Intro To Structured Credit and Equity Product PDF
Chapter 1 - Intro To Structured Credit and Equity Product PDF
Page
Introduction 1
CDS Mechanics 3
CDS Pricing 14
1
What is a Credit Default Swap?
Initial rationale:
You own a bond of company XYZ
If the company defaults you will suffer losses: Notional * (1 – R)
R: Recovery rate (price of the bond, after default, as % of notional)
How can you hedge those losses?
Selling the bond, or …
Buying a financial product which compensates you for the bond losses if default occurs
Credit Default Swaps (CDS) acts like insurance for a bond or loan – if the bond defaults then CDS
pays out (1-R) x Notional to cover bond investors’ losses.
In return for this insurance the bond holder must make regular payments.
CDS market originated in 1990s from need of banks to hedge loan risk of major clients.
INTRODUCTION
2
Agenda
Page
Introduction 1
CDS Mechanics 3
CDS Pricing 14
3
Credit Default Swap Mechanics
The protection buyer pays the protection seller regular coupons – known as the spread.
In return, if the underlying company defaults, the protection seller will make a payment to the
protection buyer equal to (1 – Recovery Rate) x Notional.
4
Credit Default Swap Mechanics
Protection Protection
buyer seller
Contingent payment on default (losses)
Payment on default =
(1 – recovery rate)
Default leg
Premium leg
CDS MECHANICS
CDS spread
CDS
(Selling
protection)
100
Risk free bond
Corporate 100
bond
CDS MECHANICS
Corporate bond holders can similarly buy CDS protection to create a synthetic “risk-free” bond.
Source: J.P. Morgan
6
Credit default swaps and cash bond exposures
Investors who only want to trade credit risk and not take exposure to government yields or swap rates
can find credit default swaps more suited to their needs.
Risk Free
7
CDS Historical Spreads: Examples
1400 300
1200 250
1000
200
800
150
600
100
400
200 50
0 0
Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
1800 800
1600 700
1400 600
1200 500
1000
400
800
300
600
CDS MECHANICS
200
400
100
200
0
0
Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
8
Source: J.P. Morgan
Example: Buying CDS protection (short risk) on Verizon
9
Example: Verizon’s credit quality worsens… its spread increases, monetise position
Assume that, after three months, Verizon’s spread widens by Investor B: Sell CDS protection
60bp to 200bp ( = 140bp + 60bp) (long credit risk)
Investor B could enter into an opposite trade, namely sell Credit: Verizon
CDS (long risk) and receive 200bp annually. They continue
to pay 140bp annually, so net 60bp per year until 2019 Notional: $10mm
Protection Buyer
Long risk for investor B (Short credit risk)
Standardised contracts
In practice:
Market uses the same “standard” contract.
Standard contract uses ISDA 2014 Credit Derivative Definitions
Standard contract terms updated in September 2014; market previously used the 2003
Definitions.
Standardisation greatly increased volumes in CDS.
CDS contracts are collateralised on a daily basis and are increasingly centrally cleared,
reducing counterparty risk.
CDS MECHANICS
11
CDS Contract Example
CDS Contract
Maturity: 5 years
12
CDS Maturities and Notional
CDS Maturities
CDS contracts are traded with standardised maturities based on IMM dates:
20th June and December of each year.
The 5y maturity is the most liquid point. Currently the active 5y maturity is the 20-Jun-2021; on 20th
September this year we will “roll” to a new contract and the 20-Dec-2021 contract will become the
new liquid 5y point.
All points in the curve are traded – 3M, 6M, 1Y, 2Y...10Y – but 75%+ volume is at 5y point.
CDS Notional
In many cases single name CDS is more liquid than corporate bonds – there can be many 5y
bonds from one company but only one standardised 5y CDS contract.
13
Agenda
Page
Introduction 1
CDS Mechanics 3
CDS Pricing 14
22
Probability Recap
Probability
To illustrate these, we play a dice game. We roll a dice five times; if the dice lands on a six we
“win” the game and do not roll the dice any more.
Cumulative Probabilities
What is the probability I win the game?
SURVIVAL AND DEFAULT PROBABILITIES
Probability of not throwing a six at all is (1 - 1/6)^5. Therefore probability of throwing at least one
six is 1 - (1 - 1/6)^5
In CDS: what is the probability the company defaults in the next five years?
Conditional Probability
What is the probability I win the game on the third roll, conditional on not having won it already?
Probability = 1/6
In CDS: what is the probability the company defaults in the third year, assuming it did not default
in the first or second year? Also known as the hazard rate.
Source: J.P. Morgan
23
Probability Recap - Continued
Probability
Unconditional Probabilities
What is the probability I win the game on the third roll of the dice?
Equal to probability of not winning game on first or second roll times probability of rolling a six
on third go.
Probability = (1 – 1/6)^2 x 1/6
In CDS: What is the overall probability of defaulting in the third year?
Summary
= PSi-1 x λi
= ( PS PS )
i 1
i 1 i
24
Survival Probabilities
We define λ as the annual conditional default probability, or hazard rate, for a particular company.
i.e. If λ for a particular company is 5%, then there is a 5% chance that this company defaults over
the next year, conditional on it not having defaulted already.
2
For payment frequency n, we can experience the survival probability at time t in years as:
t n
1
n
If we keep increasing the payment frequency, then in the limit n → ∞ this expression becomes:
t
Survival Probability at time t in years = e
This is known as a continuous payment frequency convention.
In reality we do not use a continuous payment frequency convention, but this identity is useful for
evaluating the CDS pricing equation.
SURVIVAL AND DEFAULT PROBABILITIES
26
Agenda
Page
Introduction 1
CDS Mechanics 3
CDS Pricing 14
14
Valuing a simple 1 year CDS contract
1-Year CDS
We buy protection for 1 year and will pay an amount S for this protection at the start of the year.
At the end of the year, if the company has defaulted we receive a payout of (1 – R).
In return for this protection we pay an insurance premium S at the end of the year.
For the contract to be fairly valued, S x DF1 = (1 - PS1) × DF1 × (1 - R). For interest rate r,
CDS PRICING
DFi = 1/(1+r)i
This implies that S = (1 – PS1) × (1 - R).
Source: J.P. Morgan
15
Valuing a simple 2 year CDS contract – Contingent Leg
2-Year CDS Contingent Leg
Value of Contingent Leg = (1-R) × [(1 - PS1) x DF1 + (PS1 – PS2) x DF2]
Survive
Survive 0 PS2
CDS PRICING
YEAR 1 YEAR 2
We pay the first portion of the spread S at the end of the first year as before. This acts as the
insurance premium for the first year.
We will pay the same spread S as the end of the second year as an insurance premium for the
second year, as long as the company did not default in the first year.
We only keep paying the annual spread as long as the company has not defaulted.
Default
Pay Spread
S Default
Survive
For contract to be fairly valued, value of contingent leg should equal that of fee leg.
S x DF1 + S x PS1 x DF2 = (1-R) × [(1 - PS1) x DF1 + (PS1 – PS2) x DF2]
CDS PRICING
17
How do we value a general CDS contract?
CDS Pricing
Each CDS contract has two legs: a fee leg and a contingent leg.
The fee leg is the fee (or spread) that the buyer of protection is paying.
Contingent leg
Fee leg
S S S S S S S
18
The Fee Leg
For the fee leg, spread payments will continue until the contract matures or there is a default.
The present value of the fee leg is equal to the sum of the present value of the individual spread
payments.
We need to take both the probability of survival and interest rates into account.
Fee leg
S S S S S S S
The risky present value of a single spread payment is equal to S × PSi × DFi × Δi
Maturity
Total value of fee leg = S i PSi DFi AccrualOnDefault S = CDS Spread
Psi = Probability of survival to time i
i
DFi = Discount factor at time i
Accrual On Default term accounts for any unpaid spread if default falls Δi = Length of time period i in years
between coupon periods.
19
The Contingent Leg
The present value of the contingent leg can be calculated by considering the probability of receiving
a default payout in a set period of time.
Contingent leg
S = CDS Spread
Psi = Probability of survival to time i
20
CDS Pricing
Par Spreads
A seller of protection will receive the fee leg and pay the contingent leg.
The present value of a CDS contract for a protection seller is equal to the PV of the premium leg
minus the PV of the default leg:
Maturity Maturity
PV S i PSi DFi AccrualOnDefault (1 R) ( PSi 1 PSi ) DFi
i i
A CDS contract is priced “at par” when the value of the premium leg is equal to that of the default
leg and the total PV of the CDS contract is zero.
Maturity Maturity
S i PSi DFi AccrualOnDefault (1 R) ( PSi 1 PSi ) DFi
i i
In this case the CDS spread that the seller of protection is receiving directly offsets the default risk
they are taking.
21
Valuing the CDS Fee Leg
By assuming a continuous payment convention and constant hazard rates, we can use the
Survival Probability = e-λt identity to value the CDS Fee Leg.
Maturity
1 e ( r ) T
T
1 ( r ) t
S e S
r 0 r
27
Valuing the CDS Contingent Leg
This assumes that defaults can only happen at the end of each year.
If we change our assumptions so that defaults can happen n times a year, we have:
PS PS t
lim PS t 1/ n PS t
d 1 d
lim t PS t PS t
0
dt n n dt
For
Maturity
1 d t rt Maturity
1 t rt
PV (1 R)
t 1/ n
e e (1 R) e e
n dt t 1/ n n
PV (1 R) e
T
( r ) t
dt
(1 R)
e 0 (1 R)
( r ) t T 1 e ( r )T
0 r r
Source: J.P. Morgan
28
Calculating the Par Spread
If the CDS trades at par, then the value of the fee leg is equal to that of the default leg.
For a par contract, for the Fee leg PV to equal the Default leg PV we must have S = λ x (1 – R)
This makes sense intuitively – the amount we pay for protection is equal to the payoff we expect to
receive multiplied by the probability of receiving this payoff.
30
Agenda
Page
Introduction 1
CDS Mechanics 3
CDS Pricing 14
31
Marking CDS to Market
Mark-to-market
After this, the par spread in the market changed to 200bp. This implies that the probability of default for
Verizon has changed.
Value of CDS contract for buyer of protection = PV of Contingent Leg minus PV of Fee Leg
1 e ( r ) t 1 e ( r ) t
For continuous payment convention: PVCDS (1 R) S Entry
r r
We can calculate the new probability of default from the par
spread that is currently trading in the market: SCurrent = λ(1-R)
1 e ( r ) t
Plugging this in we get: MtM CDS SCurrent S Entry
r
i.e. the mark-to-market of a CDS contract is equal to the difference in spreads multiplied by an additional
factor.
Mark to Market of CDS contract for buying protection = (Scurrent – Sentry) x Risky Annuity
Maturity
Risky Annuity = PS DF
i
i i i
Psi = Probability of survival to time i
DFi = Discount factor at time i
Δi = Length of time period i in years
32
Agenda
Page
Introduction 1
CDS Mechanics 3
CDS Pricing 14
33
Marking a CDS Trade to Market
We receive net 60bp for the remainder of the trade. Spread: 200bp
What is the mark-to-market for this trade? Maturity: 4.5 yrs (20-Jun-2021)
Buyer
(Short credit
Trade 2: 200bp Investor B2
risk)
Protection Buyer
Long risk for investor B (Short credit risk)
We need to work out what the present value of the 60bp annual payment is to calculate our mark-to-
market. We can do this by calculating the present value of a 1bp annual cashflow over the next five years
– known as an annuity.
Example: a 1bp annuity for 5y. We will receive 1bp every year for the next five years.
If interest rates and default risk is zero, the PV of this annuity is worth 5.
With non-zero interest rates, the value of later payments decreases. The new PV is 4.7.
If there is a default in the reference entity, we stop receiving the net 60bp coupon every year. Therefore,
the probability of receiving each 60bp payment is less than 100% and is linked to the riskiness of the
company.
We need to take the possibility that the company might default and we might not receive later payments
into account when calculating the PV of this annuity.
TRADING CDS IN PRACTICE
When credit risk is taken into account (the risk of not actually receiving some payments), the expected
value of later payments decreases further. The new PV is 4.5.
This PV is called the risky annuity and is very important in valuing CDS contracts.
35
Risky Annuity
PV = 300c = 3%
PV = 282c = 2.82%
TRADING CDS IN PRACTICE
PV = 270c = 2.7%
Verizon MTM
For our Verizon trade we bought protection at 140bp and sold protection at 200bp six months later.
To mark this trade to market we need to calculate the present value of this 60bp.
To calculate full P&L for this trade we also need to calculate the carry we have paid up until now.
= $70,000
37
Annuity Risk in Full Running Spread trades
Annuity Risk
We have exited our long protection position in Verizon by selling protection on the same notional.
We receive $60,000 every year from the difference in the two spreads.
The present value of these payments gives us a mark-to-market profit of $270,000
Even if the company does not default, we still have exposure to the default risk of the company. We value
our 60bp annual payment using the risky annuity; the risky annuity will fall as spreads widen and default
becomes more likely.
What’s the solution? Trading CDS on Fixed Coupons + Upfronts instead of full running spreads.
TRADING CDS IN PRACTICE
38
Fixed Coupon + Upfront Trading Convention
Since 2009, nearly all CDS contracts trade with a Fixed Coupon + Upfront format rather than Full
Running Spread.
To calculate the upfront we should pay, we need to calculate the present value of what we should be
paying (140bp) and what we are actually paying (100bp). We can do this using the risky annuity we
discussed in the previous section.
If the quoted spread is higher than the coupon then the buyer of protection will pay the upfront.
TRADING CDS IN PRACTICE
If the quoted spread is lower than the coupon then the seller of protection will pay the upfront.
For our Verizon example, the buyer of protection will pay $180,000 (=4.5 x (140-100) x $10mm)
39
Fixed Coupon + Upfront Trading Convention
Going back to our valuation of the CDS fee and contingent legs, the upfront for a fixed coupon trade is
equal to the PV difference, when S is equal to the fixed coupon.
Maturity Maturity
Upfront (1 R) ( PSi 1 PSi ) DFi C i PSi DFi AccrualOnDefault
i i
1 e ( r )T 1 e ( r )T
Upfront (1 R) C
r r
If we take the current par spread as S = λ(1 – R) then we obtain
1 e ( r )T
Upfront ( S C )
r
TRADING CDS IN PRACTICE
1 e ( r )T
Where S is the par spread, C is the fixed coupon and is the risky annuity for flat curves and a
continuous payment convention. r
40
Trading Verizon with Fixed Coupons and Upfronts
We carry out the same Verizon trade as before, but use an upfront + fixed coupon format rather than a
full running spread format this time.
Six months later, we sell protection at 200bp with a fixed coupon of 100bp on notional of $10mm.
Upfront = 4.0 x (200bp – 100bp) x $10mm = $400k.
We receive upfront of $400k and receive coupon payments of $100k every year.
We have paid six months of carry equal to $50,000 (= 100bp x 0.5 x $10mm).
TRADING CDS IN PRACTICE
41
Annuity Risk in Fixed Coupon + Upfront Trades
If the company defaults we have no exposure as we are not receiving any net payments.
There is no annuity risk in Fixed Coupon + Upfront trades – as long as trades use the same fixed
coupon.
TRADING CDS IN PRACTICE
42
Risky Annuity and Risky Duration (DV01)
Another often used number is the Risky Duration (or just duration).
Risky duration – the change in mark-to-market for a 1bp change in spread.
For Fixed Coupon + Upfront trades, the Risky Duration is just equal to the difference in upfronts for a 1bp
change in spread.
Upfront for spread S and coupon C = RAS x (S – C)
Upfront for spread S+1 and coupon C = RAS+1 x (S + 1 – C)
Remember – risky annuity changes as spread changes.
Risky Duration ≈ Risky Annuity – many market participants use the two terms interchangeably.
The bigger difference between the spread and the coupon, the bigger the difference between risky
duration and risky annuity.
43
Agenda
Page
Introduction 1
CDS Mechanics 3
CDS Pricing 14
44
Additional Material
Additional Material
CDS v2.0: The new architecture of the CDS market, 2010, J.P. Morgan
Understanding CDS Upfronts, Unwinds and Annuity Risk, 2008, J.P. Morgan
45
CDS CURVES
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
1
CDS Curves: Credit Risk for Different Maturities
CDS Curves
Credit default swaps on the same company can have different spreads for different maturities.
The entire term structure of CDS spreads is known as the “CDS Curve”.
1600
1400
1200
1000
800
600
400
200
0
1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y
11-Dec-2012 26-Oct-2012
INTRODUCTION
Source: J. P. Morgan
2
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
3
CDS Curves
CDS Curves
Average annual probability of default over i years = Par Spread of CDS contract with maturity i / (1 minus Recovery)
Some companies are very safe for the next few years and very risky after that – simply looking at the 5y spread will not tell
you this.
CDS curves give us information about how the default risk of a company changes over time
4
How do we calculate Hazard Rates?
Hazard Rates
Hazard rates are conditional probabilities of default; i.e. what is the probability that a company defaults in year 3, given that it
has not defaulted in years 1 or 2?
Based on an annual payment convention, we can calculate the probability of survival to year i using the Par CDS Spread to
that date: i
S
PSi 1 i (1)
1 R
We can also calculate the probability of survival to year i using hazard rates:
PSi 1 j
i
(2)
j 1
By setting these two equations equal to one another, we can calculate a hazard rate λi provided that we know the hazard
rates λ1,..., λi-1 and the i-year Par CDS Spread Si. This gives us:
1
1 S i
(1 R)
i
(3)
1
i i 1
j
j 1
This equation allows us to use a method called bootstrapping to calculate hazard rates from a CDS curve.
5
Bootstrapping Hazard Rates
We assume that we know the CDS Par Spread Si for any maturity date i from observing levels trading in the market.
We make an assumption about the recovery rate R, usually equal to 40% for senior CDS and 20% for sub CDS.
Firstly, the average annual probability of default we calculate from the 1Y CDS spread must be equal to the 1 st year hazard
rate, given that we assume the company has not yet defaulted as of today.
S1
1 (4)
1 R
From the equation on the previous page, we can gain an expression for λ2 in terms of S2, λ1 and R.
1
1 S (1 R)
2
2
(5)
2
1 1
Once we have calculated a value for this expression, we know the values of both λ1 and λ2. Using the 3-year Par CDS
Spread we can continue this method and calculate λ3:
1
1 S (1 R)
3
3
1 1
3 (6)
1 2
We can continue this method to calculate values for hazard rates λ4, λ5, λ6... and so on.
6
Bootstrapping Hazard Rates
Given (1) and (2), it is also possible for us to write a general formula for the ith year hazard rate λi using only market traded
CDS spreads and an assumed recovery rate:
PSi
i 1 (7)
PSi 1
i 1
1 S i
(1 R)
i
(8)
1 S i 1
(1 R)
i 1
7
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
8
Flat Curves
Flat CDS Curves CDS Spread Curves
Names with flat curves also have flat hazard rate curves. 1000
800
The riskiness of the name is not changing with time – it has
an equal probability of defaulting in each of the next 10 600
years. 400
CDS Spreads
20% 6%
4%
10%
2%
0%
1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y 0%
1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y
Cumulative Probability of Default
Hazard Rates
9
Upward Sloping Curves
Upward Sloping CDS Curves CDS Spread Curves
CDS Spreads
Source: J.P. Morgan
20% 6%
4%
10%
2%
0%
1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y 0%
1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y
Cumulative Probability of Default
Hazard Rates
10
Downward Sloping Curves
Downward Sloping CDS Curves CDS Spread Curves
The first few years are the most risky, with annual default 600
risk falling over time. 400
CDS Spreads
Names with inverted curves are often close to default.
Source: J.P. Morgan
20% 4%
10% 2%
0% 0%
1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y 1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y
Cumulative Probability of Default Hazard Rates
11
Comparison
Comparison of different curves CDS Spread Curves
30% 6%
20% 4%
10% 2%
0% 0%
1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y 1Y 2Y 3Y 4Y 5Y 6Y 7Y 8Y 9Y 10Y
12
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
13
Carry and Slide
14
Carry and Slide for Full Running Spread Trades
We sell 5y protection on a full running spread format on a name with the following par CDS curve:
Tenor Spread Risky Annuity
1Y 80bp 0.9
2Y 95bp 1.8
3Y 110bp 2.7
4Y 130bp 3.6
5Y 150bp 4.5
Every year we will receive a coupon of 150bp. Carry over next 12m = 5y Spread = 1.50%
After one year we expect the 5Y spread to roll down to 130bp. The slide is equal to the mark-to-market resulting from this
spread move:
Slide over next 12m= 4y Risky Annuity x (5y Spread – 4y Spread) = 3.6 x (150bp – 130bp) = 0.72%
If we had bought 5y protection the rolldown in the curve would work against us, so the time value would be -2.22%.
15
Carry and Slide for Fixed Coupon + Upfront Trades
Carry, Slide and Time Value for Fixed Coupon + Upfront Trades
We sell 5y protection on a fixed coupon plus upfront format on a name with the following par CDS curve with a coupon of
100bp: Tenor Spread Risky Annuity
4Y 130bp 3.6
5Y 150bp 4.5
Every year we will receive a coupon of 100bp. Carry over next 12m = Fixed Coupon = 1.00%. Note this is different to the
Full Running Spread case, where the carry was equal to the 5y Spread.
One year later we expect the upfront we unwind the trade at to be equal to:
4y Upfront = 4y Risky Annuity x (4y Spread – Coupon) = 3.6 x (130bp – 100bp) = 1.08%
The upfront has decreased by 1.17% over the year, resulting in a mark-to-market gain of 1.17% in our favour. Some of this
change in upfront is due to the change in maturity and some is due to the change in spread; we want to separate the two
impacts.
CARRY AND SLIDE FOR CDS POSITIONS
If the spread had stayed constant at 150bp, the upfront would have still have changed due to the shorter maturity. We assume
that the 4y risky annuity would be equal to 3.5 if this spread had remained constant. This would have resulted in a new 4y
upfront of 1.75%. This means the upfront would have reduced by 0.50% even if the spread had remained constant. We call
this the Excess Carry. Excess Carry over next 12m = 0.50%.
The slide is the change in the upfront not explained by the excess carry.
Slide over next 12m = 5y Upfront – 4y Upfront – Excess Carry = 2.25% - 1.08% - 0.50% = 0.67%
Time Value = Carry + Excess Carry + Slide = 1.00% + 0.50% + 0.67% = 2.17%
Note that the combined Carry and Excess Carry are very close to the Carry from the Full Running Spread Example
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
17
Exposures for Curve Trades
In a curve trade an investor will buy CDS protection on one part of the curve and sell protection on another part.
e.g. Buy 5y protection and Sell 10y protection.
In this sort of trade there are three main P&L sensitivities that investors will be concerned with:
Exposure to moves in spreads – how will the trade perform in various spread scenarios?
Time value exposure – how much money do I expect the trade to make over time if the curve shape remains constant?
Default exposure – how will the trade react to defaults?
These sensitivities are dependent on the relative notionals used in the trade. There are two main notional scalings that
investors use:
Equal Notional. Each leg has the same notional.
Duration-Weighted. Longer maturity legs have lower notionals to offset the higher duration.
Terminology:
Buy long dated protection, sell short dated protection: Steepener
Sell long dated protection, buy short dated protection: Flattener
18
Spread Exposure I
For a full running trade the P&L impact on a 3s5s flattener from spread moves is equal to:
S
2
The term arises due to changes in duration as spreads change.
In a duration weighted trade, an investor chooses the notionals such that the trade is not exposed to parallel shifts in spreads.
We can solve for these notionals by setting P&L = 0 and S3Y S5Y
Duration5 y
This gives the following notional relationship for duration-weighted trades: Notional3 y Notional5 y
Duration3 y
Source: J.P. Morgan
P&L FOR CURVE TRADES
19
Spread Exposure II
Spread Exposure
When looking at the sensitivities of curve trades we look at both the exposure to spread changes and how the P&L changes with
time.
Curve Risk
P&L FOR CURVE TRADES
Key takeaway: Equal notional trades have both spread and curve risk, duration-weighted trades only have curve risk
20
Carry, Slide and Time Value for Curve Trades
Spread Exposure
Similar to single name CDS trades, investors entering curve trades will often calculate the carry, slide and time value of their
positions in order to try and project future P&L behaviour.
Calculating carry, slide and time value for a curve trade is straightforward; we use the same calculation as for a single name
CDS position but then combine the values for the two legs.
We enter a 3s5s equal notional steepener on ABC Corp on a notional of €10,000,000 on a full running spread format.
Tenor Spread Duration
1Y 120 0.9
2Y 140 1.8
3Y 160 2.7
4Y 180 3.6
5Y 200 4.5
3Y Leg
Carry = 160bp x 1 year x €10,000,000 = €160,000
Slide = (160bp-140bp) x 1.8 x €10,000,000 = €36,000
Time Value = +€196,000 for 3y leg
Total Time Value = Time Value for 3y Leg + Time Value for 5y Leg = €196,000 - €272,000 = -€76,000
Curve trades can have varying types of jump to default exposure. The profit from defaults for a 3s5s flattener is equal to:
Equal notional trades usually have minimal default exposure due to two legs cancelling each other out.
The exact default exposure will depend on the relative recovery rates for the two legs.
For Bankruptcy, Failure to Pay and Governmental Intervention credit events the recoveries will be equal for all maturities.
For Restructuring events shorter dated contracts may have higher recoveries.
22
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
23
Why Trade Curves?
Duration Weighted Trades let investors take a view on the timing of defaults without taking outright exposure to spread risk.
Duration-weighted steepeners express a view that defaults are likely to occur later than the market is pricing in.
Duration-weighted flatteners express a view that defaults are likely to earlier later than the market is pricing in.
If I buy 5y protection and sell 3y protection, effectively I am buying protection starting in 3 years and ending in 5 years.
Equal notional steepeners express a view that defaults are likely to occur in a specific period.
Equal notional steepeners express a view that default risk is likely to be lower in a specific period than the market is pricing in.
24
Forward CDS
Equal notional curve trades are effectively buying or selling forward-starting CDS protection.
For example, if I enter a 3s5s equal notional flattener (buy 3y protection, sell 5y protection in equal notionals) I am effectively
selling 2y protection starting in 3 years.
Forward starting contracts do not trade by themselves in the market; you have to construct them with an equal notional curve
trade.
We can calculate an implied spread for a forward starting CDS contract by considering the value of the fee leg for a short-
dated contract with maturity i, a long dated contract with maturity j and a forward starting contract covering the period
between i and j.
The value of the fee leg for any contract is equal to the risky annuity RA times the par spread S.
Therefore:
RAi Si RAi , j Si , j RA j S j (9)
Where RAj and Sj are the risky annuity and spread of a contract starting at i and ending at j.
We can re-arrange this to give an expression for the implied forward spread Sj:
RA j S j RAi Si
Si , j (10)
RAi , j
WHY TRADE CURVES?
25
Forward CDS
Given that:
i
RAi k PSk DFk (10)
k
Then we deduce that the Risky Annuity of a contract starting at i and ending at j is just equal to the difference in the risky
annuities for maturities i and j:
This allows us to rewrite the equation for the implied CDS forward spread purely in terms of spreads and risky annuities of
conventional CDS contracts:
RA j S j RAi Si
Si , j (12)
RA j RAi
Example: ABC Corp
5y S=100bp RA=4.5
10y S=150bp RA=8.0
26
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
27
CDS Curve Behaviour – Tight Spread Names
this.
28
CDS Curve Behaviour – Distressed Names
29
Curves inverted during the 2008-2009 Crisis
150
100
50
0
-50
-100
-150
-200
-250
-300
HISTORICAL CDS CURVES
Jan-07 Apr-07 Jul-07 Oct-07 Jan-08 Apr-08 Jul-08 Oct-08 Jan-09 Apr-09
3s5s 5s10s
30
Curves have steepened since 2010
20
This steepening trend has been driven by a number of
factors: 10
The current steep level means that duration-weighted Xover 3s5s and 5s10s Curves
flatteners are now positive time value for 3s5s curves in 150
Main and Crossover.
100
This provides a strong technical against further 50
steepening; investors will often look to add these trades as
0
portfolio hedges when they offer positive time value.
-50
-100
HISTORICAL CDS CURVES
Jan-10 Apr-10 Jul-10 Oct-10 Jan-11 Apr-11 Jul-11 Oct-11 Jan-12 Apr-12 Jul-12 Oct-12
3s5s 5s10s
31
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
32
Risky Annuities
From the last lecture, the general formula for risky annuity is: T = Maturity (in years)
1T n n = Payment frequency
Risky Annuity = PSi DFi PSi = Survival probability at i
i 1 n DFi = Discount factor at i
For a continuous payment convention and a constant hazard and interest rate we showed that the risky annuity is equal to:
1 e T ( r )
T = Maturity (in years)
λ = Annual default probability
r = Interest rate
r
However, as we have discussed in this lecture, annual default probabilities are not constant with time.
Fully calculating the Risky Annuity requires us to taking changing default probabilities into account.
Under an annual payment convention, the risky annuity for a 5-year CDS contract can be expressed as:
PS1 DF1 PS2 DF2 PS3 DF3 PS4 DF4 PS5 DF5
If the CDS spread for maturity i is Si and the average annual default probability implied by this spread is λi then the i-year
probability of survival is equal to PSi = (1 - λi )i.
33
Flat Spreads
Nearly all CDS trades today are done on a fixed coupon + upfront convention. RA = Risky Annuity
S = Quoted CDS Spread
The CDS spread is just a quotation convention – we never actually pay or receive the spread. C = Coupon
What we do need is a clear, standardised way of getting from a quoted spread to a traded upfront which all counterparties
agree on,
We need the risky annuity to calculate the upfront, which is equal to: Upfront RAS C
To calculate the risky annuity we need to know the entire CDS curve; this presents a problem, as the counterparty we are
trading with might agree on a 5y spread with us, but if we disagree on the 1y, 2y, 3y or 4y spread we are going to calculate a
different risky annuity and disagree on the upfront.
Importantly, when we trade on a fixed spread + upfront convention, the quoted spread is a “flat spread” and will be different to
the par spread, due to the assumption of a flat curve affecting the risky annuity.
The flat spread can be higher or lower than the par spread; this depends on the shape of the curve and the coupon used.
FLAT SPREADS
34
A Flat Spread Example
ABC Corp trades with a par 5y CDS spread of 150bp and the curve is upwards sloping.
160
We want to do a trade with a fixed coupon of 100bp.
140
120
Using the par spread, the upfront of this trade is equal to:
100
Upfront RAPar S Par C 80
60
We want to know what flat spread we should quote for this trade; i.e. the flat 40
20
curve which gives us an upfront equal to that we calculate for the upward 0
sloping curve. 1Y 2Y 3Y 4Y 5Y
Par Spread Coupon
Based on a recovery of 40%, we calculate RAPar as 4.72. This gives us an upfront
of 2.36% based on the par spread curve (see below). We assume interest rates are zero.
If we assume a flat curve equal to the 5y spread of 150bp, RAFlat = 4.64. This is lower than the Risky Annuity for the upward
sloping curve and so the upfront is lower at 2.32%.
We need to shift the flat curve upwards to match the flat spread-generated upfront to the par spread curve generated upfront.
If we shift the flat curve to 151bp, this gives us a new RAFlat = 4.63 and an upfront of 2.36%.
The 5y Flat Spread for this curve is 151bp, 1bp higher than the par spread of 150bp.
35
Agenda
Page
Introduction 1
Hazard Rates 3
Flat Spreads 32
CDS CURVES
36
Additional Material
Additional Material
Trading CDS with Fixed Coupons: How it affects basis and curve positions, 2010, J.P. Morgan.
37
BOND-CDS BASIS
Agenda
Page
Introduction 1
1
Comparing Cash Bonds and CDS
Corporate Bonds and Credit Default Swaps both reference the credit risk of the same company.
Investors can choose to trade this credit risk through either bonds or CDS.
However, sometimes the two different products give different prices for this credit risk. This is known as the Bond-CDS Basis.
2
The Bond-CDS Basis
The Bond-CDS Basis measures the difference in the bond credit spread and the CDS spread.
If the basis is positive, CDS are trading wide to bonds: CDS cheap to cash bonds
If the basis is negative, CDS are trading tight to bonds: CDS expensive to cash bonds.
Investors can trade bonds and CDS together to take advantage of the bond-CDS basis.
Negative basis trade: Buy bonds and buy CDS protection when basis is negative
Positive basis trade: Sell bonds and sell CDS protection when basis is positive.
However, differences in bond and CDS markets mean that these packages are not true arbitrages.
3
Agenda
Page
Introduction 1
4
Which bond spread to use?
Bond spreads
To calculate the bond-CDS basis we need both a bond spread and a CDS spread.
CDS spread is relatively straightforward – we look up the spread of a CDS contract with a maturity matched to that of the bond
(linearly interpolating if necessary).
It’s not so clear which bond spread we should use. The most common choices are:
Z-Spread. This is one of the most commonly used bond spreads with many bonds quoted on a z-spread basis. However,
z-spreads are not actually traded in the market – I cannot enter a trade which pays me the z-spread.
Asset swap spread. Less commonly quoted than z-spreads, but the advantage is that you can actually receive the asset
swap spread by entering an asset swap package.
Each of these has advantages/disadvantages; we also use the Par Equivalent CDS Spread (PECS).
5
Par Equivalent CDS Spread
PECS
The Par Equivalent CDS Spread is a JPM-developed measure designed to calculate a bond spread using the same pricing
techniques we use to price a CDS.
This gives a credit spread measure which is as comparable as possible with a CDS spread.
PECS Calculation:
The CDS curve implies survival probabilities for the company at each point in the future.
We can apply these survival probabilities and the assumed recovery rate to the future cashflows of the bonds:
This gives us a “CDS-implied” price for the bond as follows:
Maturiy
CDS Implied Bond Price CPS DF 100 PS
i 1
i i Maturity
DFMaturity
This CDS-implied price will be different to the bond price we see in the market; we apply a shift to the CDS curve until the
CDS-implied price matches the bond price seen in the market.
When the two prices match, the shifted CDS spread with a maturity equal to the bond maturity is the PECS.
MEASURING THE BASIS
Asset swap spreads and Z-spreads are useful and quick measures to calculate the basis, but they are not designed to be
compared against a CDS spread. In most cases the PECS will be similar to the z-spread and asset swap spread, but there
can be a large difference for distressed names.
6
Agenda
Page
Introduction 1
7
Trading the Basis
Basis Trades
If the bond-CDS basis is above or below zero, then we can look to enter trades to take advantage of this pricing difference.
If basis is negative:
CDS spread is tighter than bond spread... i.e. Buying CDS protection is cheap.
Negative basis trade: Buy bond and buy CDS protection.
If basis is positive:
CDS spread is wider than bond spread... i.e. Buying CDS protection is expensive.
Positive basis trade: Sell bond and sell CDS protection.
8
Trading the basis – Negative Basis Trades
When the basis is negative, the cost of protection is less than the excess credit spread we get from owning the bond.
If the company defaults, we can deliver the bond into the CDS contract and receive par.
Rate of Return from holding a negative basis trade: = Risk free rate + bond spread – CDS spread
= Risk free rate - bond CDS basis
P&L from holding basis trade ≈ (Exit Basis – Entry Basis) x Duration - Entry Basis x Days/360 – Funding Costs
However, trade is not “risk-free” – there are potential pitfalls with negative basis trades.
Very important to check deliverability criteria – is bond deliverable into CDS contract?
9
Negative Basis Trade P&L
For a negative basis trade, we expect to receive an annual return equal to the risk-free rate minus the bond-CDS basis.
e.g. If the risk free rate is 50bp and the bond-CDS basis is -40bp, the average return of the negative basis trade will be
90bp (=50bp – (-40bp)).
This is like a bond yield – it represents the expected average return to maturity but in reality the annual payments can be
structured differently.
To get the whole picture, we need to consider the cost we pay to enter the negative basis package and what happens if there
is a default.
e.g. if the bond price is 90 and the CDS upfront for a buyer of protection is 5%, then the upfront cost of entering a negative
basis trade is 95 (= 90 + 5).
In general, investors prefer basis trades which have a basis package cost of less than 100. If the cost is above 100, then an
immediate default will result in a loss as the investor only receives 100 back from the CDS payout.
If the basis package cost is below 100, then an immediate default benefits the investor.
10
Negative Basis Trades – worked example
We buy bond at 95pts and buy CDS protection at 120bp (upfront of 1.18pt).
Risk free rate = 135bp. Average annual return = 165bp (= 135 + 30bp).
Going forward we receive annual net coupon of 1%; we also receive the benefit of the package trading at less than par.
If there is no default between now and maturity, total P&L is equal to 1% x 6 years + (100 – 96.18) = 9.82%.
If there is an immediate default this trade does well because the package costs less than 100.
Total P&L from basis trade based on timing of default Average annual return based on timing of default
TRADING THE BASIS
SNS Bank NV
SNS Bank subordinated debt was expropriated by the Dutch government in February 2013
Owners of sub debt found that they no longer owned the subordinated debt and they had received zero in return.
SNS bond holders could not deliver their bonds as they did not own them, Sub CDS contracts paid out very little.
Negative Basis Trades on SNS Bank sub debt ended very badly
12
Trading the basis – Positive Basis Trades
When the basis is positive, the CDS spread trades wide compared to the bond spread.
Some investors try to take advantage of this by shorting bonds and selling CDS protection.
P&L from holding basis trade ≈ (Entry Basis – Exit Basis) x Duration + Entry Basis x Days/360 – Funding Costs
Generally, positive basis trades are much more difficult to execute than negative basis trades due to difficulties in shorting
bonds.
Even if it is not possible to execute a positive basis trade, the basis still gives you a useful signal about relative value.
If the basis is positive then selling CDS protection can be more attractive than buying the cash bond.
TRADING THE BASIS
13
Agenda
Page
Introduction 1
14
Why does the basis exist?
Bond-CDS Basis
Investors are not always able to take advantage of technical reasons via positive/negative basis trades due to funding costs or
other constraints, meaning the basis can persist for some time.
Call features
A bond may be expected to be called early, meaning that the spread is tighter than the CDS to the maturity date.
15
Why does the basis exist?
Deliverability issues
The bond may be issued from a different entity than the one the CDS references. If the bond entity is not guaranteed and is of
lesser credit quality than the CDS will trade tighter than the bond.
Risk of orphaning
Bond tenders can lead to orphaned entities, meaning that CDS contracts trade on a reference entity with no deliverable debt.
This can lead to negative basis on bonds which are likely to be tendered.
Funding costs
Entering negative basis packages requires cash; if investors have to borrow cash to enter basis packages they will have to
pay the funding costs associated with this. This means that investors will only enter negative basis trades when the basis is
negative enough to offset these funding costs.
Rating constraints
Many IG bond investors are forced to sell bonds if the company is downgraded to HY. This can cause underperformance of
the bonds relative to HY in the run up to and aftermath of a downgrade.
Bond issuance
WHY DOES THE BASIS EXIST?
Heavy supply in bond markets can cause bonds to move wider relative to CDS.
16
Agenda
Page
Introduction 1
17
Historical case study – 2008 and 2009
stabilised.
18
Historical case study – 2012 and 2013
19
The € IG Bond-CDS Basis is currently positive in Europe
0%
-80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80
70 70
60 60
50
50
40
THE HISTORICAL BOND-CDS BASIS
30 40
20 30
10
0 20
-10 10
-20
-30 0
Apr-13 Jun-13 Aug-13 Oct-13 Dec-13 Feb-14 <-400 -350 -300 -250 -200 -150 -100 -50 0 50 100 150 200 250 300 350 >350
20
Agenda
Page
Introduction 1
21
Additional Material
Additional Material
Trading with Fixed Coupons: How it effects basis and curve positions, 2010, J.P. Morgan
22
CDS INDICES
Agenda
Page
Index Mechanics 11
Basis to Theoretical 15
1
What is a CDS Index?
CDS Indices
e.g. iTraxx Main is a basket of 125 single name CDS contracts in European Investment Grade.
CDS indices trade as a single contract, meaning that you can get macro exposure to credit without trading a large number of
single name CDS contracts.
CDS indices are usually equally weighted; each name in iTraxx Main makes up 1/125th or 0.8% of the index notional.
The cost of buying protection on iTraxx Main will be very similar to the cost of buying protection on all 125 single names, but
there can be some differences.
2
Agenda
Page
Index Mechanics 11
Basis to Theoretical 15
3
Global CDS Index Overview
European CDS Indices: Sovereign Indices
iTraxx Europe Family iTraxx SovX
iTraxx Europe SovX Western Europe
ITraxx Main
SovX CEEMEA
+ Senior / Sub Financials SovX Asia
ITraxx HiVol Asia CDS Indices:
ITraxx Crossover (High yield) iTraxx Asia Family
iTraxx Asia
US CDS Indices:
ITraxx Japan
CDX Family
CDX ITraxx Asia Ex-Jp IG & HY
DJ CDX.NA.IG
ITraxx Australia
DJ CDX.NA.HY
Emerging Markets CDS
Indices: CDX EM
GLOBAL CDS INDICES
CDX EM
DJ CDX.EM
4
European CDS Indices
3y, 5y, 7y and 10y points are traded (5y most liquid). Trades on upfront + 500bp fixed coupon.
5
iTraxx Main and Crossover Spreads
250
200
150
100
50
0
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
iTraxx Main
Source: J.P. Morgan
1400
1200
1000
800
600
400
200
GLOBAL CDS INDICES
0
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
iTraxx Crossover
Source: J.P. Morgan
6
European CDS Indices - continued
30 European Banks and Insurance Companies 30 European Banks and Insurance Companies
iTraxx Senior Fins references the same 25 Financials that Same 25 names as Senior Financials
appear in iTraxx Main.
References the Lower Tier 2 debt
References the Senior Unsecured debt.
5y and 10y points are traded (5y most liquid).
5y and 10y points are traded (5y most liquid).
Quoted on spread.
Quoted on spread.
Trades on upfront + 500bp fixed coupon.
Trades on upfront + 100bp fixed coupon.
€1bn daily volume across entire market.
€6bn daily volume across entire market.
GLOBAL CDS INDICES
7
iTraxx Senior and Subordinated Financials Spreads
700
600
500
400
300
200
100
0
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
iTraxx Senior Fins iTraxx Sub Fins
Source: J.P. Morgan
Sub/Senior Ratio
2.2
2
1.8
1.6
1.4
1.2
GLOBAL CDS INDICES
1
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
Sub/Senior Ratio
Source: J.P. Morgan
8
US CDS Indices
125 North American Investment Grade Companies 100 North American High Yield Companies
Constituents must be rated higher than BBB-/Baa3 Neg Constituents must be rated BBB-/Baa3 Neg Outlook or
Outlook. below.
Composition is based on the names with the highest Composition is based on the names with the highest
trading volumes. trading volumes.
3y, 5y, 7y and 10y points are traded (5y most liquid). 3y, 5y, 7y and 10y points are traded (5y most liquid).
9
CDX IG and CDX HY Spreads
CDX IG Spreads
300
250
200
150
100
50
0
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
CDX IG
Source: J.P. Morgan
CDX HY Spreads
2000
1500
1000
500
GLOBAL CDS INDICES
0
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
CDX HY
Source: J.P. Morgan
10
Agenda
Page
Index Mechanics 11
Basis to Theoretical 15
11
CDS Index Mechanics
CDS indices are quoted and traded in the same way as single name CDS.
Example:
We buy iTraxx Main protection on €100million for a 5y maturity.
Quoted spread is 55bp
We will pay the fixed coupon of 100bp.
Due to the difference in the quoted spread and fixed coupon we will also receive an upfront amount, equal to 2.025 % of
notional.
12
CDS Index Series
A new series is launched every six months on 20th March and 20th September – this is known as the “index roll”.
The new series of each index has a new composition and six months extra maturity.
The current “on-the-run” series is Series 21 for iTraxx indices and Series 22 for CDX indices.
For example, on 20th March 2014 we rolled from iTraxx S20 to iTraxx S21.
iTraxx S21 indices had an updated composition reflecting liquidity and rating changes since 20th September
iTraxx S21 5y maturity is 20-Jun-2019, 6 months longer than the iTraxx S20 5y maturity of 20-Dec-2018.
Old indices become “off-the-run”. They still trade but become less liquid over time.
13
CDS Indices and Credit Events
The defaulted single name is not replaced; the CDS index The triggered name is stripped out of the contract and
continues to trade on a reduced notional. settled separately like a normal single name CDS.
Example:
We buy €125mm of iTraxx Main protection (125 names),
equivalent to €1mm protection on each of the single
names. The protection buyer will deliver the bonds and
There is a credit event on one name with a 40% receive par (like a normal cds)
recovery.
We receive €600k = €1mm x (1-40%) from the single
name portion of the index. The contract remains live on a reduced notional
The index will continue to trade and will now be known
as “Version 2” of the Series. The new reduced notional
of the index will be €124mm.
Premium is paid or received on a reduced notional
INDEX MECHANICS
14
Agenda
Page
Index Mechanics 11
Basis to Theoretical 15
15
Traded vs. Theoretical Spread: Basis to Theoretical
Basis to theoretical
We can also calculate a theoretical spread for the index using the spreads of the underlying single names.
These spreads will often not be the same – the difference is known as the basis-to-theoretical.
If the basis gets too large, investors will look to arbitrage the difference.
16
Trading the Basis
Basis to theoretical
The index and underlying single names hold exactly the same default risk.
If the basis-to-theoretical is very positive or negative, investors will look to arbitrage the difference by trading both the index
and all the underlying single names.
Example:
iTraxx Main is trading at a spread of 100bp and the underlying single names give a theoretical spread of 90bp.
Basis to theoretical = +10bp.
We sell protection on iTraxx Main at 100bp and buy protection on all 125 single names.
To make a trade worthwhile the basis needs to be large enough to offset the cost of the bid-offer on the index and single
names.
Arb lists in general are difficult to execute, given the problems associated with trying to line up a large number of single name
trades simultaneously.
17
Calculating the Theoretical Index Spread
Theoretical Spread
We can calculate a theoretical index spread using single name trading levels.
As a starting point, the upfront of the index should be equal to the sum of the upfronts of all the single names, adjusted for
their weighting within the index. N
U I = ∑ ωiU i
UI = Index Upfront
Ui = Single Name Upfront
ωi = Weighting of name in index
i =1 N = Number of names in index
If the index is equally weighted (as most are), then ωi = 1/N. We can also express the upfronts in terms of risky annuities,
spreads and coupons. SI = Index Spread
Si = Single Name Spread
1 N
RAI ( S I − C ) = ∑ RAi ( S i − C )
RAI = Index Risky Annuity
RAi = Single Name Risky Annuity
N i =1 C = Fixed Coupon
1 N
We can rewrite the index risky annuity RAI as the sum of the individual single name risky annuities RAI = ∑ RAi
N i =1
This gives us: SI N C N 1 N C N
∑ RAi − ∑ RAi = ∑ RAi S i − ∑ RAi
N i =1 N i =1 N i =1 N i =1
N
∑ RA S i i
The coupons cancel out and we can then express SI as SI = i =1
N
∑ RA
i =1
i
BASIS TO THEORETICAL
Wider names have lower risky annuities and therefore a lower weighting – they are more likely to default so we expect to pay
the spread for a shorter amount of time.
The lower weighting of wider spreads means that the Theoretical Spread is lower than a simple average of the single name
spreads.
Source: J.P. Morgan
18
Historical Basis-to-Theoretical
iTraxx Main
20
10
0
-10
-20
-30
-40
-50
-60
-70
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
iTraxx Main
iTraxx Crossover
100
50
-50
-100
BASIS TO THEORETICAL
-150
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10 Jan-11 Jul-11 Jan-12 Jul-12 Jan-13 Jul-13
iTraxx Crossover
19
Agenda
Page
Index Mechanics 11
Basis to Theoretical 15
20
Additional Material
Additional Material
21
INTRODUCING CDS INDEX OPTIONS
Agenda
Page
Intro 1
Option Strategies 34
INTRODUCING CDS INDEX OPTIONS
1
Market Structure Update
Trades by Counterparty
€500bn annual option volume
2
CDS Index Options
Option characteristics
Traded on 5y index for all of the above as well as 10y index for iTraxx Main.
Option expiries generally traded up until one year, though longer expiries may be available on request.
CDS index options are European options; can only be exercised on expiry date between 9am and 4pm.
Options can also be unwound/offset before expiry to take profit or cut losses.
3
CDS Index Options
Option types:
Payer: Option to buy protection, i.e. pay coupons
Receiver: Option to sell protection, i.e. receive coupons
The option buyer pays an upfront premium which is usually denominated in cents; 1 cent = 0.01% on an upfront basis.
Example:
Client buys iTraxx Main payer option with expiry of 18-Jun-14 and strike of 120bp for 75c on notional of €100mm.
If the spread had been lower than 120bp, the contract would not have been exercised and would have expired worthless.
4
CDS Example Run
Index Series Expiry
Forward Current Spread
Prices
5
Receiver & Payer Options
Large upside when spreads tighten Unlimited upside when spreads widen
Downside limited to premium paid Downside limited to premium paid
Increasing spreads
Increasing spreads
Large downside when spreads tighten Unlimited downside when spreads widen
Upside limited to premium earned Upside limited to premium earned
INTRO
6
Expressing a bullish view on spreads
tightening
Decreasing price Linear return profile if spreads widen or
Increasing spreads
tighten
20bp 50bp 80bp
Unlimited downside risk if spreads widen
Increasing spreads
7
Expressing a bearish view on spreads
widening
Decreasing price Linear return profile if spreads widen or
Increasing spreads
tighten
20bp 50bp 80bp
Downside risk capped as spreads cannot
be negative
Page
Intro 1
Option Strategies 34
INTRODUCING CDS INDEX OPTIONS
9
CDS Option Questions
At what spread level do we breakeven, taking the cost of the option into account?
10
A simple example
We buy a Crossover payer option with an expiry of 18-Jun-14 and a strike of 500bp on €100mm notional.
We have the right to buy iTraxx Crossover protection at 500bp on 18-Jun-14. We will only do this if the spread on this date is
greater than 500bp.
Initial Cost
11
A simple example - breakevens
If spreads are above 500bp at expiry, we will exercise our option and make a profit from selling index protection higher than
the 500bp level at which we bought protection.
However we need to make back the cost we paid for the option - this means we need the index to widen to more than 500bp
to make an overall profit.
This option will only make a total profit if spreads are wider
than 543bp at expiry.
12
A simple example – P&L at expiry
We can work out our total P&L at expiry in the same way we calculated the breakeven spread.
If we exercise the payer option, we will buy protection at the strike K. We will then mark this protection to market at the current
spread S.
P&L if we exercise the option = (Spread – Strike) x Forward Duration x Notional – Option Cost
In previous example, we exercise the option when spreads are 600bp and the forward duration is 3.5
13
Agenda
Page
Intro 1
Option Strategies 34
INTRODUCING CDS INDEX OPTIONS
14
What happens if there is a default in the index?
This means that options give you protection on defaults before expiry.
For example, if we buy a December payer option and there is a default in October, at expiry we receive the same default
payout we would have received from an index position as long as we exercise the option.
This means that the buyer of a payer option has default protection starting today, even though they are only buying
protection starting on the exercise date.
Example: we buy a Crossover payer option with a December expiry on a notional of €100mm.
Codere has a failure to pay credit event on 15th September with a recovery rate of 40%.
At expiry we exercise the option. We receive a payout of €1.2mm from the Codere event (= €100mm/50 x (1-40%)) and also
get entered into Version 2 of the index the specified strike.
INDEX DEFAULTS AND FORWARD SPREADS
15
Forward Spreads
This is the market-implied expectation for where spreads will be at the option expiry.
The conventional CDS forward is calculated from the CDS curve; steeper curves mean higher forwards.
If option expires at time E and index matures at time T, then CDS forward from expiry to maturity is equal to:
ST RAT − S E RAE
Forward =
RAT − RAE
where Si is the index spread to time i and RAi is the risky annuity to time i.
The no-knockout convention means that payer options are worth more than they would be otherwise. This is
incorporated into the option price by making an adjustment to the forward level.
INDEX DEFAULTS AND FORWARD SPREADS
This “no-knockout adjustment” is equal to the carry we would have paid for protection between now and the option expiry,
annualised over the life of the forward contract:
S E RAE
NoKnockoutAdjustment =
RAT − RAE
The adjusted forward is equal to the conventional CDS forward plus the no-knockout adjustment.
16
The Adjusted Forward
Adjusted Forward
The Adjusted Forward is equal to the current spot level plus the spot times the risky annuity to expiry, divided by the forward
risky annuity. This is roughly equal to:
17
Agenda
Page
Intro 1
Option Strategies 34
INTRODUCING CDS INDEX OPTIONS
18
Option Valuation
CDS index option prices are calculated using a Black-Scholes framework, modified for the CDS market.
Pricing inputs:
Current spreads
Strike
Expiry
Volatility
Rates
Options mean you get all the upside and none of the downside from market moves.
Compare two different CDS indices, both of which have spreads of 100bp today.
Index A moves up or down by 1bp each day.
Index B moves up or down by 10bp each day.
If we could buy a 100bp strike payer option on Index A or Index B for the same price, we would always buy the Index B
option.
We don’t care whether spreads tighten by 1bp or 10bp – our option is worthless in either case.
We make much more money if spreads widen by 10bp than 1bp.
CDS OPTION VALUATION
We expect to make more money from a payer option on Index B; the price of this option will therefore be higher.
19
Implied and Realised Volatilities
We can back out equivalent implied volatilities from the traded prices.
To get an idea of whether these implied volatilities are cheap or expensive, we can compare them to historical realised volatilities.
There are a number of different ways to calculate realised volatility. The most commonly used realised volatility is equal to:
∑ [ln(S S )]
n
2
Realised vol =
i i −1
i =1
× 252
n −1
Comparing implied and realised volatilities gives us an idea of whether now is a good or bad time to buy or sell options. For example,
if implied volatility was much higher than recent realised, this may be a good time to sell options.
CDS OPTION VALUATION
Remember though – realised volatility is a backward looking measure, implied volatility is a forward looking measure.
20
CDS Option Pricing
Option Pricing
If we exercise the option, there will be an upfront amount to pay/receive corresponding to the strike.
U K = RAK ( K − C )
Upon exercise we will be entered into a long protection position on the index. If the current market spread is S then the
upfront of this position will be equal to:
U S = RAS ( S − C )
The MtM gain, or option value, at expiry due to the exercise of the option is equal to the difference in these two upfronts.
MtM S , K = U S − U K
If we know the value of the option at expiry as a function of the market spread, we can write the value of the option today as
the following: ∞
where DFE is the discount factor to expiry, P(S) is the probability distribution of the index spread S at the expiry of the option.
The standard is to assume a lognormal distribution with volatility σ and mean F, where F is the adjusted forward spread.
[ln( S / F )]
2
1
P( S ) = exp −
CDS OPTION VALUATION
S 2π σ 2σ 2
This equation can then be solved in a similar fashion to the Black-76 method.
This equation is an approximation (and neglects some more detailed parts of default handling) but gives a very good
indication of CDS option pricing.
Source: J.P. Morgan
21
Valuing options before expiry
Option Value Option Value (cents) for iTraxx Main 100bp 3M Payer
Using the equation set out on the previous page we can 250
150
An option will always have a positive value prior to expiry,
even if it is unlikely to be exercised. 100
50
Investors used partial derivatives of the option price
equation to keep track of how the price of the option is 0
60 70 80 90 100 110 120 130 140
affected by changing markets.
Today Expiry
These are collectively known as the Greeks
Source: J.P. Morgan
Delta. Change in the option price with respect to a
change in the index spread. Option Value (cents) for iTraxx Main 100bp 3M Receiver
Gamma. Change in delta with respect to a change in
250
the index spread.
Theta. Change in option price with respect to time. 200
Today Expiry
22
Delta
Option Delta Option Delta (%) for iTraxx Main 100bp 3M Payer
i.e. if index duration is 4.5, then if index widens by 1bp then 20%
index upfront increases by 4.5c. 0%
60 66 72 78 84 90 96 102 108 114 120 126 132 138
If option price increases by 2.25c for this 1bp move in index Today Expiry
then delta = 50% (=2.25/4.5).
∂POption ∂U Index 1 ∂POption Source: J.P. Morgan
Delta = =
∂S I ∂S I DurationIndex ∂S I Option Delta (%) for iTraxx Main 100bp 3M Receiver
0%
For a payer option, the delta before expiry increases as the
-20%
spread widens; the option becomes more sensitive to
spread moves as it comes more in-the-money. -40%
-60%
At expiry, delta = 0% for an out-of-the-money option and
-80%
100% for an in-the-money option.
-100%
60 66 72 78 84 90 96 102 108 114 120 126 132 138
CDS OPTION VALUATION
Today Expiry
23
Gamma
Option Gamma Option Gamma (%) for iTraxx Main 100bp 3M Payer
spread. 5%
4%
This can be seen as the convexity of the option position. 3%
2%
For constant duration:
1%
∂Delta 1 ∂ 2 POption 0%
Gamma = = 60 66 72 78 84 90 96 102 108 114 120 126 132 138
∂S I DurationIndex ∂S I2 Today 1 week before expiry
In general, higher gamma is preferable – a more convex Option Gamma (%) for iTraxx Main 100bp 3M Receiver
position makes gains bigger and losses smaller.
6%
Gamma is highest for ATM options and becomes very 5%
peaked when options approach expiry. 4%
3%
Short term ATM options have very high gamma.
2%
Short term ITM or OTM options have low gamma. 1%
0%
Gamma is directly proportional to theta – see next slide. 60 66 72 78 84 90 96 102 108 114 120 126 132 138
CDS OPTION VALUATION
24
Theta
Option Theta Option Theta (c) for iTraxx Main 100bp 3M Payer
For example, if theta is -3c, we expect the option price Source: J.P. Morgan
gamma. -1.0
60 66 72 78 84 90 96 102 108 114 120 126 132 138
CDS OPTION VALUATION
25
Vega
Option Vega Option Vega (c) for iTraxx Main 100bp 3M Payer
1.2
Vega measures the change in option price with respect to
1.0
changes in implied volatility.
∂POption 0.8
Vega = 0.6
∂σ 0.4
0.2
Convention is to define theta as the change in the option 0.0
price (in cents) for a 1% increase in implied volatility if 60 66 72 78 84 90 96 102 108 114 120 126 132 138
Today 1 week before expiry
spreads are kept constant.
For example, if vega is 4c, we expect the option price to Source: J.P. Morgan
26
Delta Hedging
For most options the biggest sensitivity is the delta; i.e. the 2,500,000
2,000,000
sensitivity to changes in the index spread. 1,500,000
1,000,000
However, it is possible to negate this exposure by trading 500,000
a certain amount of index. 0
-500,000
Example: -1,000,000
-1,500,000
We buy a 100bp strike payer option with a delta of 50% 60 68 76 84 92 100 108 116 124 132 140
on a notional of €100mm. We are short risk – we will Buy €100mm Payer Option Sell €50mm Index
make money if spreads go wider. Source: J.P. Morgan
If index duration is 4.5, then for a 1bp move wider in
index spreads the option price will increase by 2.25c
and the value of the position will increase by €22,500 = Delta Hedged Position – Index + Option
(50% x 4.5 x €100mm). 800,000
If we sell €50mm of index protection, then for a 1bp 700,000
600,000
move wider we will lose €22,500 (=4.5 x €50mm).
500,000
The index position cancels out the option position; the 400,000
sensitivity to small index moves has been removed. 300,000
200,000
This is known as Delta Hedging and allows investors to 100,000
0
take direct views on volatility without taking spread risk. 60 68 76 84 92 100 108 116 124 132 140
CDS OPTION VALUATION
28
Trading Gamma vs. Theta
29
Trading Vega
30
Agenda
Page
Intro 1
Option Strategies 34
INTRODUCING CDS INDEX OPTIONS
31
Historical implied vs. realised volatility
Selling Credit Volatility P&L from selling short dated vol on iTraxx Main since 2008
As a result selling volatility is now much less profitable iTraxx Main – Implied vs. Realised Volatility
than it was previously.
HISTORICAL OPTIONS VOLATILITIES
Source: J. P. Morgan
Source: J. P. Morgan
32
Volatility skew
Source: J. P. Morgan
Source: J. P. Morgan
33
Agenda
Page
Intro 1
Option Strategies 34
INTRODUCING CDS INDEX OPTIONS
34
Option Strategies – Range Strategies
Source: J. P. Morgan
Butterflies Strangles
OPTION STRATEGIES
Source: J. P. Morgan
Source: J. P. Morgan
35
Option Strategies – Bullish Strategies
Bullish risk reversal: buy low strike receiver, sell high strike
payer. Benefits from steep skew.
Source: J. P. Morgan
Source: J. P. Morgan
Source: J. P. Morgan
36
Option Strategies – Bearish Strategies
Payer spread: buy low strike payer, sell high strike payer.
Benefits from steep skew.
Payer ladder: buy low strike payer, sell 2 x high strike payer.
Benefits from steep skew.
Bearish risk reversal: buy high strike payer, sell low strike
receiver. Benefits from flat skew.
Source: J. P. Morgan
Source: J. P. Morgan
Source: J. P. Morgan
37
Agenda
Page
Intro 1
Option Strategies 34
INTRODUCING CDS INDEX OPTIONS
38
Additional Material
Additional Material
39
CDS TRANCHES AND CORRELATION TRADING
Agenda
Page
What is a tranche? 1
Tranche pricing 11
Tranche sensitivities 21
1
Recap – Losses for CDS Baskets
2
CDS baskets only offer a single spread
Investors like selling protection on CDS baskets and indices and it diversifies risk.
The expected P&L volatility on a 125-name basket with an average spread of 100bp is much less than that of selling
protection on one single name with a spread of 100bp.
The downside is that investors have different return targets and risk appetities and the basket spread might not meet these.
Example 1: Hedge Fund investor aims for an annual return of 5% on a portfolio of €100million.
To achieve this return, the Hedge Fund would have to sell €500million of iTraxx Main protection.
The downside of this is that the maximum loss is €500million; more than the value of the Hedge Fund’s portfolio.
Example 2: Pension Fund has an annual return target of 0.5% on a portfolio of €100million.
To achieve this return, the Pension Fund would have to sell €50million of iTraxx Main protection.
The downside is that the Pension Fund is still exposed to defaults in the index, despite their lower risk appetite.
In short – standard baskets such as iTraxx indices may be liquid and diversify risk, but they do not offer a one-size-fits-all
solution to the market. Investor demand is much more varied and diverse.
3
What is a tranche?
4
Different tranches will appeal to different investors.
Tranches
Let’s go back to our previous example – iTraxx Main trades at 100bp and two clients (a Hedge Fund and a Pension Fund) want
to sell protection to meet their annual return targets.
A 0-10% tranche on iTraxx Main trades at 500bp and a 10-100% tranche trades at 55bp.
This pricing is arbitrary for now – we will discuss more on pricing later.
The advantage of this is that the maximum loss on the 0-10% tranche position is €100million; to achieve the same return on the
index the maximum loss would have been €500million.
The 0-10% tranche offers what is known as non-recourse leverage; investors can achieve a higher return on the same portfolio
of names without increasing their maximum loss.
The advantage of this is that the pension fund no longer takes losses for the first few defaults in the index, which they would
have done if they had sold €50million of index protection at 100bp.
Tranches allow investors to choose an investment with a risk profile which suits them, from the same portfolio of names.
WHAT IS A TRANCHE?
5
Bespoke CSOs and Index Tranches
Bespoke CSOs
Prior to the 2008-2009 Financial Crisis, investors and banks would regularly structure highly-customised tranches, using a
portfolio of names and risk level selected by the client.
For example, a client might ask a bank to structure a 7-8% tranche on a portfolio chosen specifically by the client.
This market largely shut down in mid-2008 with the onset of the Financial Crisis.
There has been some re-emergence of the bespoke CSO market in recent years, but new regulation makes it very capital
intensive for banks to structure these products.
Index Tranches
These standardised tranches were originally used by banks to hedge bespoke CSOs they had structured.
CDS index tranches are the most liquid structured credit product today.
6
Index Tranches
22-100% 31bp
Equity
iTraxx Main Index 92bp
7
Step-by-step: how do defaults affect tranches?
For a 40% recovery, each default in iTraxx Main results in an index loss of 0.48% (= (1-40%)/125).
What impact do defaults have on the 0-3% and 3-6% tranche? Defaults Index Loss 0-3% Loss 3-6% Loss
0 0.00% 0% 0%
First default causes a 16% notional loss to the 0-3% tranche (=0.48%/3%); 1 0.48% 16% 0%
the seller of 0-3% protection will pay out 16% of the tranche notional. 2 0.96% 32% 0%
3 1.44% 48% 0%
By the time of the 7th default, the 0-3% tranche has been wiped out and the 4 1.92% 64% 0%
3-6% tranche starts absorbing losses. 5 2.40% 80% 0%
6 2.88% 96% 0%
Each subsequent default causes a 16% loss to the 3-6% tranche, until the 7 3.36% 100% 12%
13th default when the 3-6% tranche is wiped out. 8 3.84% 100% 28%
9 4.32% 100% 44%
The lower the recovery, the larger the loss that each default causes to a 10 4.80% 100% 60%
tranche. 11 5.28% 100% 76%
12 5.76% 100% 92%
13 6.24% 100% 100%
8
The Current iTraxx tranche market.
iTraxx S9 tranches
The most liquid iTraxx tranche markets are current based on iTraxx Main S24 and S9.
The legacy iTraxx tranche market is based on the iTraxx Main Series 9 portfolio.
9
Trading and Quoting Formats for iTraxx Tranches
Like CDS indices, tranches are traded on an upfront + fixed coupon format.
10
Agenda
Page
What is a tranche? 1
Tranche pricing 11
Tranche sensitivities 21
11
Tranche Pricing
Tranche Pricing
The loss of tranche j at time t as a % of the tranche notional is given by Z j ,t and can be expressed as:
min (Z t , K D , j ) − min (Z t , K A, j )
Z j ,t =
K D , j − K A, j
where Zt is the index loss as a % of index notional at time t and K A , j and K D , j are the attachment and detachment
points of tranche j respectively.
To calculate the expected loss on tranche j we need to know the probability distribution of Zt
Once we have the expected loss, we can calculate the spread and/or upfront of the tranche.
12
Tranche Pricing - continued
Tranche Pricing
= ∫ Z j ,t Ρ(Z t )dZ t
1
Expected loss of tranche j at time t
0
=
K D, j − K A, j
This cannot be inferred from the spread of the index, or the spreads of individual single names in the index.
14
Index Loss Density Function P(Zt)
Distribution of Index A Distribution of Index B
12%
30%
10% Y-Axis: P(Zt) Y-Axis: P(Zt)
25%
X-Axis: Zt X-Axis: Zt
8%
20%
6%
15%
4%
10%
2%
5%
0%
0%
-2%
0% 20% 40% 60% 80% 100% -5%
0% 20% 40% 60% 80% 100%
The two above charts show two different distributions with the same expected loss of 15% - i.e. the mean of the distribution is
equal to 15%.
If the overall expected loss is the same, then the two indices should trade at the same spread.
However, the different distributions will have a large impact upon tranche pricing within the index.
For example. a senior tranche on A will trade with a very tight spread; there is very little chance of the index loss exceeding
20%.
However, on index B there is a chance, albeit a small one, of a very large number of defaults happening at once (signified by
the small hump to the right side of the Index B x-axis). This means there is some chance that the senior tranche will take
significant losses, so we expect the senior tranche on Index B to trade wider than that on Index A.
TRANCHE PRICING
Correlation is another important input into tranche pricing, in the same way volatility is an input into option pricing.
15
Agenda
Page
What is a tranche? 1
Tranche pricing 11
Tranche sensitivities 21
16
Correlation and Expected Loss
Tranche Pricing
In a low correlation environment, defaults are isolated with no real knock on effects or common factors.
In a high correlation environment, defaults occur in quick succession and are often linked by some common factor.
For tranches, people often look at the expected loss of a portfolio. This is the total amount of loss we expect for a portfolio
over its lifetime.
Expected Loss = Spread x Risky Annuity
iTraxx Main S9 Expected Loss = 4.1%
THE SIGNIFICANCE OF CORRELATION IN TRANCHES
17
Why is correlation important?
Tranche Pricing
In the first portfolio, there is a 50% chance of a 5% portfolio loss and a 50% chance of a 15% portfolio loss.
Expected Index Loss = 10%
If there is a 5% loss, the 0-10% tranche loses 50% of notional and the 10-100% tranche loses 0% of notional.
If there is a 15% loss, the 0-10% tranche loses 100% of notional and the 10-100% tranches loses 5.6% of notional.
– Expected loss for 0-10% tranche = 75%
– Expected loss for 10-100% tranche = 2.78%
In the second portfolio, there is a 90% chance of a 0% portfolio loss and a 10% chance of a 100% portfolio loss.
THE SIGNIFICANCE OF CORRELATION IN TRANCHES
The first portfolio is an example of a low correlation portfolio, whereas the second portfolio is a high correlation portfolio.
The total amount of risk is the same in both portfolios; the expected loss of each index is the same so the two indices will
trade at the same spread.
Correlation affects the distribution of risk between equity and senior tranches; senior tranches are riskier when correlation
is high.
18
A correlation analogy
Tranche Pricing
Imagine a minefield with a set number of mines. There are five lanes through which travel is possible.
While the number of mines is constant, the distribution is not, as shown below.
The tank is the “senior tranche” and the bicycle is the “equity tranche”.
THE SIGNIFICANCE OF CORRELATION IN TRANCHES
19
Dispersion and Implied Correlation
Correlation Types
When we speak about correlation, we are actually describing two different factors with similar effects.
1. Dispersion
This is effectively the correlation already priced into a portfolio. A highly dispersed portfolio will have some names trading at
very wide spreads with some names trading very tight. A portfolio with low dispersion will have nearly all names trading close
to the average spread.
A portfolio with high dispersion decreases the correlation – equity tranches underperform and senior tranches outperform
in a high dispersion portfolio.
A portfolio with low dispersion increases the correlation – equity tranches outperform and senior tranches underperform in
a low dispersion portfolio.
2. Implied Correlation
THE SIGNIFICANCE OF CORRELATION IN TRANCHES
This is what the tranche pricing is implying for correlation in the future. In a similar way to how implied volatility is backed out
from options pricing, the implied correlation is backed out from tranche prices once factors like portfolio risk and dispersion
have been taken into account.
A portfolio with high implied correlation leads to wider senior tranches and tighter equity tranches.
A portfolio with low implied correlation leads to tighter senior tranches and wider equity tranches.
Dispersion is an input to the model; implied correlations are backed out from market traded prices in the same way that
implied volatilities are backed out from options prices.
20
Agenda
Page
What is a tranche? 1
Tranche pricing 11
Tranche sensitivities 21
21
Tranche sensitivities
Sensitivities
How does the price of each tranche change as these factors change?
22
Tranche Delta
Delta
The total default risk across all tranches must be the same as the default risk in the index.
If the index spread widens, the spreads of each tranche will widen as well.
The more junior the tranche, the higher the sensitivity of the tranche to changes in the index spread.
We define this as the % mark-to-market change in the tranche divided by the % mark-to-market change in the index for a 1bp
move in the index.
For example, if the index widens by 1bp and has a duration of 4, the MtM change in the index is 0.04%. If the tranche upfront
changes by 0.20% then the tranche delta is 5 (=0.20%/0.04%).
TRANCHE SENSITIVITIES
23
Correlation Exposure
Correlation sensitivity
For mezzanine tranches, the exposure to correlation depends on the tranche’s location relative to the expected loss of the
portfolio.
In general:
When the tranche is below the expected loss, the tranche will act like an equity tranche and will widen when correlation
falls.
When the tranche is above the expected loss, the tranche will act like a senior tranche and will widen when correlation
increases.
For example, if the portfolio expected loss is 4%, we expect the 9-12% tranche to act like a senior tranche and
underperform when correlation increases.
Table below shows exposure to 1% correlation increase for S9 tranches, both for outright and delta-adjusted notionals.
Current S9 expected loss is 4.1%; 0-3% and 3-6% tranches are long correlation while other tranches are short.
TRANCHE SENSITIVITIES
24
Default Exposure
Default sensitivity
Equity tranches have the biggest sensitivity to defaults given they will be impaired from the very first default.
Other tranches have mark-to-market exposure; a 3-6% tranche is riskier after one default than it is after zero defaults; the
extra default has effectively causes the tranche to become more subordinated.
25
Time Value
We see the same effect with tranches, but the rate of time
decay depends very much on the tranche in question.
Mezzanine tranche time value relative to index Senior tranche time value relative to index
TRANCHE SENSITIVITIES
26
Delta Hedging Tranches
If we want to trade correlation with no exposure to underlying spreads, we can remove a tranche’s inherent exposure to the
index spread by trading the underlying index.
For example, if we sell protection on the equity tranche (with a delta of 7) with a notional of €10mm we can buy €70mm of
index protection against this to hedge out the exposure to the index.
If we think correlation is going to increase, we can position for this in a number of ways, including:
Selling equity protection and delta hedging.
Buying senior protection and delta hedging.
If correlation does increase, then the equity tranche should tighten and the senior tranche should widen relative to other
tranches.
Source: J.P. Morgan
TRANCHE SENSITIVITIES
27
Delta Hedged Equity Tranches
0 -8,000,000
0m 6m 12m 18m 24m -600,000 0 5 10 15 20
-6% -4% -2% 0% 2% 4% 6%
Source: J.P. Morgan Source: J.P. Morgan Source: J.P. Morgan
28
Delta Hedged Senior Tranches
we are: -5,000
-10,000
Negative convexity -15,000
Negative time value -20,000
-600,000
-2,000,000
-300,000 -800,000 0 5 10 15 20
0m 6m 12m 18m 24m -6% -4% -2% 0% 2% 4% 6%
Source: J.P. Morgan Source: J.P. Morgan Source: J.P. Morgan
29
Agenda
Page
What is a tranche? 1
Tranche pricing 11
Tranche sensitivities 21
30
The 2005 Correlation Crisis
by S&P. 20%
31
The 2008-2009 Financial Crisis
rose; the financial crisis affected all companies, not just an 20%
10%
isolated few as was the case with GM and Ford.
0%
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11 Jan-12 Jan-13
0-3% correlation went from around 10% in early 2007 to
0-3% Correlation
more than 50% in late 2008.
Source: J.P. Morgan
This increase in correlation caused a large
underperformance of senior tranches. The 10y 22-100%
iTraxx Main 22-100% Spread
tranche went from trading at around 5bp in early 2007 to
wider than 70bp at the peak of the crisis. 100
HISTORICAL TRANCHE PRICING AND THEMES
90
80
70
60
50
40
30
20
10
0
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11 Jan-12 Jan-13
22-100% Spread
Source: J.P. Morgan Source: J.P. Morgan
32
A Historical Perspective
Structured Credit
CDS tranches are sometimes referred to as Collateralised Synthetic Obligations (CSOs), especially for bespoke baskets of
CDS single names.
Prior to the 2008-2009 financial crisis, tranching portfolios of risk was very popular and there was very high demand for the
junior (i.e. equity and mezzanine) tranches given their high returns.
There was less demand for the senior tranches; many banks were left holding these senior tranches on their balance sheets.
When correlation rose during the financial crisis, these senior tranches performed very badly and in many cases were written
down with very large losses.
HISTORICAL TRANCHE PRICING AND THEMES
The worst performing senior tranches were those with underlying portfolios based on subprime mortgages.
33
Correlation since 2009
34
Agenda
Page
What is a tranche? 1
Tranche pricing 11
Tranche sensitivities 21
35
A Model for Tranche Pricing
Tranche Pricing
As discussed in the previous section, a core problem in tranche pricing is to calculate the probability density function of the
portfolio loss Zt
Ρ( Z t ) was given by Oldrich Vasicek in a 1987 paper. This is known as the Vasicek model.
An initial approach to calculating
This model makes a number of assumptions which allows us to calculate a closed-form solution for Ρ ( Z t )
To start off we give an overview of a Merton structural model for credit risk:
At time t, firm n has assets worth An ,t and debt outstanding of D with maturity T.
n
dAn ,t
We assume that the value of An ,t follows a continuous-time diffusion: = rdt + σ n dWn ,t
An ,t
where r is the risk-free rate,σ n is the asset volatility and Wn ,t is a standard Brownian motion.
Simplistically, a firm can be judged to be in default when the value of its assets is less than the outstanding value of its debt.
APPENDIX I: THE VASICEK MODEL
Based on this, the probability at time t of a firm defaulting at the maturity of the debt T, conditional on the assets at time t
being equal to An ,t , is equal to:
pn ,t ,T = Ρ( An ,T < Dn | An ,t ) (1)
36
The Merton Structural Credit Model
Tranche Pricing
Based on (1), we can then use Itô’s lemma to express the asset value at time T as a function of the current asset value as
follows:
σ n2
An ,T = An ,t exp r − (T − t ) + σ n T − t X n ,t ,T (2)
2
Where X n ,t ,T is given by:
Wn ,T − Wn ,t
X n ,t ,T = ~ N (0,1) (3)
T −t
Therefore, at time t, we can express the condition for firm n defaulting at the debt maturity T using two separate conditions:
where:
σ
2
ln Dn − ln An ,t − r − n (T − t )
K n ,t ,T = 2 (5)
σn T −t
APPENDIX I: THE VASICEK MODEL
Given that X n ,t ,T ~ N (0,1) , the probability at time t of firm n defaulting at time T can be written as:
pn ,t ,T = Ρ( X n ,t ,T < K n ,t ,T ) = Φ (K n ,t ,T ) (6)
37
The Vasicek Model
Vasicek I
The Vasicek model builds on the Merton Structural Model by combining single name default probabilities into a probability
density function for the portfolio loss on a portfolio of names.
For an index of N names we need to somehow account for correlations between individual names.
Firstly, we re-use X n ,t ~ N (0,1) from the Merton model, which is the a random variable determining the status (default
or not default) of name n at time t.
We assume that the correlation coefficient of each pair of random variables X n ,t and X m ,t is ρ n ,m ,t .
The first major assumption within the Vasicek model is to assume that the correlation between every pair of single names is
equal.
38
The Vasicek Model
Vasicek II
We can think of the random variable driving firm’s assets values as being attributable to two factors: an idiosyncratic factor
affecting only that name and a systemic factor which affects all names. The magnitude of the correlation ρ t determines the
relative importance of each factor.
X n ,t = ρ t Yt + 1 − ρ t ε n ,t (8)
Where Yt , ε 1,t ,..., ε N ,t are independent standard normal random variables. We can interpret Yt as representing the
systemic factor which affects all names equally, and ε n ,t as an idiosyncratic factor affecting only firm n. As ρ t increases the
systemic factor becomes more important and the idiosyncratic factor less important.
Conditional on knowing the systemic risk factor Yt at time t, we write the probability of default for firm n as:
pn (Yt ) = Ρ( X n ,t < K n ,t | Yt ) (9)
(
= Ρ ρ t Yt + 1 − ρ t ε n ,t | Yt ) (10)
K − ρ t Yt
APPENDIX I: THE VASICEK MODEL
39
The Vasicek Model
Vasicek III
Assumption II: We know the individual default probabilities of each firm: p1,t ,..., p N ,t
We can calculate the default probabilities implied by single name CDS spreads trading in the market.
If this is the case, then the inverse of (6) gives us K n ,t = Φ −1 ( pn ,t ) . Substituting this into (12) gives:
Φ −1 ( pn ,t ) − ρ t Yt
pn ,t = Φ
(13)
1 − ρt
Assumption III: The default probability of all firms is the same and equal to pt
This is a major assumption, but it greatly simplifies the model and for the sake of brevity of algebra we will assume it here.
Extending the model to non-constant default probabilities is relatively straightforward.
pt = Φ
(14)
1 − ρ t
For each firm n, we define a random variable Ln ,t which takes the following values:
40
The Vasicek Model
Vasicek IV
1 N
Ωt = ∑ Ln ,t (14)
N i =1
The cumulative distribution function of Ωt is equal to:
F (ω ; pt , ρ t ) = Ρ(Ω t ≤ ω ) (15)
The index loss,Z t , which we are attempting to calculate a probability density function for, is given by:
1 N
Zt = ∑ Ln ,t (1 − Rn ,t ) (16)
N i =1
where Rn ,t is the CDS recovery rate for firm n at time t.
Assumption IV: The CDS recovery rate Rn ,t is constant through time and equal to R for all firms.
APPENDIX I: THE VASICEK MODEL
This implies:
(1 − R ) N
Zt = ∑ L = (1 − R)Ω
n ,t t
(17)
N i =1
41
The Vasicek Model
Vasicek V
Based on (15) and (17), we can write the cumulative distribution function of Z t as:
ω
Ψ (ω ; pt , ρ t ) = F
(18)
; pt , ρ t
1 − R
Assumption VI: The number of namesN in the portfolio is very large such thatN →∞
When we apply the law of large numbers to the portfolio, we find that the fraction of defaulted credits in the portfolio Ω t tends
to the individual default probability p (Yt ) ; this is fully explained in Schönburger (2000) and Vasicek (2002).
As a result:
F (ω ; pt , ρ t ) = Ρ( p (Yt ) ≤ ω ) (19)
Φ −1 ( pt ) − ρ t Yt
= Ρ Φ < ω
(20)
1 − ρt
Φ −1 ( pt ) − 1 − ρ t Φ −1 (ω )
= Ρ Yt ≥
(21)
ρt
APPENDIX I: THE VASICEK MODEL
1 − ρ t Φ −1 (ω ) − Φ −1 ( pt )
= Φ
(22)
ρ t
42
The Vasicek Model
Vasicek VI
Combining (18) and (22) gives the following expression for the cumulative distribution function of Zt :
1 − ρ t Φ −1 (ω (1 − R) ) − Φ −1 ( pt )
Ψ (ω ; pt , ρ t ) = Φ
(23)
ρt
d
Ρ (Z t = ω ) = Ψ (ω ; pt , ρ t ) (24)
dω
P(Zt) for pt = 40%, R = 40%, ρ = 40%. P(Zt) for pt = 40%, R = 40%, ρ = 70%.
APPENDIX I: THE VASICEK MODEL
0% 10% 20% 30% 40% 50% 60% 0% 10% 20% 30% 40% 50% 60%
43
Base Correlations
Using the Vasicek model we can back out implied 450 No solutions above 400bp
400
correlations from tranche spreads and prices trading in the 350
market. 300
250
However, this presents a problem for mezzanine tranches; 200
150
for many spreads there are two implied correlations which
100 Two solutions
could apply. 50 for 250bp
0
This issue does not apply for equity or senior tranches. -50
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
As a result, most modern tranche models use Base
Correlations.
tranches.
e.g. if we wanted to buy protection on a 3-6% tranche,
we could buy 0-6% protection and sell 0-3% protection.
44
Agenda
Page
What is a tranche? 1
Tranche pricing 11
Tranche sensitivities 21
45
Additional Material
Additional Material
CDS Index Tranche Strategies: What can tranches do for you?, 2011, J.P. Morgan
Credit Risk Models IV: Understanding and pricing CDOs, 2005, Abel Elizalde
46
March 2016
Davide SilvestriniAC
davide.silvestrini@jpmorgan.com
+44(0) 20 7134 4082
This presentation was prepared exclusively for instructional purposes only, it is for your information only. It
is not intended as investment research. Please refer to disclaimers at back of presentation.
Equity Dividend Swaps
Gaining pure exposure to dividends
Mechanics
Pricing: Put-Call parity & dividends, futures & dividends, pull to realised
Drivers
1
Dividends
Dividends are payments that companies makes to their shareholders, typically to
distribute part of profits.
2
Dividends and Equity Derivatives
The fair value of most equity derivatives depends on expected dividends.
For example, the fair value of equity futures is based on expected dividends
FT = S ⋅ e rT − E [ Div ]
Exception: derivatives on Total Return Equity indices (i.e. indices where dividends
are reinvested) and Total Return Equity Swaps have no exposure to dividends
Dividends represent a risk for equity derivatives market makers. The fair value of
their derivatives position will change if dividend expectations turn out to be incorrect
or even if dividend expectations change (mark-to-market risk)
Dividend derivatives were created as a mean to trade dividend risk in isolation, for
improved risk management and/or speculation
SWAPS
Dividend swaps
Dividend futures
DIVIDEND
Options on dividends
3
Dividend Swap
A dividend swap offers straightforward and direct exposure to the dividends paid
by an underlying stock or index.
Realised
dividends
Dividend Dividend
Seller Buyer
Implied (“agreed”)
dividends
We first introduce dividend swaps for single stocks. We then cover dividend swaps for
stock indices.
DIVIDEND
4
Mechanics of a dividend swap
Swap legs & counterparties
Swap dates
Mark-to-Market
5
Mechanics: Swap legs and counterparties
Dividends / Swap Legs:
Realised (floating leg): total amount of all qualifying dividends going ex by the
underlying equity or index, between the start date (exclusive) and end date
(inclusive).
For single stocks, “realised” and “implied” dividends refer to dividends per share.
If the stock or index delivers (i.e. “realises”) a higher dividend amount than the
fixed leg of the swap, the long dividend position will make a profit. Conversely,
SWAPS
if the stock or index delivers a lower dividend amount, then the long dividend
swap position will suffer a corresponding loss.
DIVIDEND
If realised dividends turn out to be equal to the implied dividend, the payoff of
the swap at expiry is zero.
6
Mechanics: Swap dates I
Important dates in a dividend swap:
Start date: date from which “qualifying” dividends will be accumulated in the
floating leg of the swap.
Dividend swaps are generally quoted for one particular calendar year, with the end-
dates corresponding to the listed option expiry date closest to end of a calendar year
and the start-dates set at the previous contract’s expiry.
In order to align with normal exchange expiry schedules, the dates chosen are
commonly the third Friday of December.
Example: 2014 dividend swap included all dividends going ex between
SWAPS
7
Mechanics: Swap dates II
Dividend swaps are quoted & traded as “forward” instruments.
For example, the 2016 dividend swap includes only dividend payments between
Dec-2015 and Dec-2016.
How can you construct a dividend swap which includes dividend payments from
the trade date (inception) to Dec-16?
8
Mechanics: Swap Notional & Payout
Dividend swaps are cash settled at the swap expiry date.
The notional size of the dividend swap is also necessary to calculate final P&L. For
single stocks this is given by the number of shares.
Final payoff (at expiry) at expiry is equal to the difference between the
realised and implied dividend multiplied by the number of shares specified in
the dividend swap.
Example 1:
SWAPS
An investor thinks the dividend paid by XYZ LtD in 2014 is at risk of being lower than the €0.5 currently implied (i.e.
traded). To express this view, the investor sells a Dec-14 dividend swap with a “notional” of 2,000,000 shares.
XYZ announces a dividend of €0.4 due to go ex-dividend on the 24th July. No other dividends are paid.
DIVIDEND
Thus PnL at expiry (i.e. December) to the short position = -2,000,000 * (€0.4-€0.5) = +€200,000
What will be the mark-to-market of the trade following the ex-date?
9
Mechanics: “Qualifying” (realised) dividends (I)
The payout of a dividend swap is based upon realised dividends that go ex-
dividend between the start and expiry dates.
The date that matters for dividend derivatives is the ex-dividend date, not the
dividends announcement date or payment date.
The gross dividend amount is typically used in dividend swaps, i.e. before any
deduction or credit for withholding tax, stamp tax, or any other tax, duties,
fees or commissions.
10
Mechanics: “Qualifying” dividends (II)
The derivatives exchange or index provider defines which dividends qualify as ordinary
dividends. Regular dividends are most likely recurring, paid out of operating earnings, and
typically are not a large percentage of the current share price. Dividends not fitting this pattern
are considered special or extraordinary and are not included.
Dividends are usually counted before any deduction or credit for withholding tax, stamp tax,
or any other tax, duties, fees or commissions.
Cash dividends paid as a result of a return of capital or reduction of par value may or may
not be included in full. The exact treatment of return of capital dividends depends on the index
sponsor.
Dividends paid in stock may or may not be included. Stock dividend paid with treasury shares
are included (they are not dilutive). Scrip dividends where the shareholder has the option to
either receive cash of shares also qualify.
For example, Telefonica pays scrip dividends, which count in full for dividend swaps
despite the fact that most shareholders choose to receive shares.
SWAPS
If in doubt contact the exchange or index calculation agent. Dividend swaps are generally
designed to align with exchange options and forwards. Most exchanges publish exact rules for
DIVIDEND
11
Mechanics: Mark-to-Market
Example:
An investor thinks the dividend paid by XYZ LtD in 2016 is at risk of being lower than the €0.5 currently being
implied (i.e. traded). To express this view, the investor sells a Dec-16 dividend swap with a “notional” of 2,000,000
shares.
The implied dividend falls to €0.4 the next day.
What is the MtM on this position?
The dividend swap payout is valued at the expiry of the dividend swap and so the
SWAPS
and has no discounting (this is due to the daily margining of the position).
12
Dividend: Swaps vs. Futures
Dividend Swap are OTC products
Differences:
As OTC products, the terms in dividend swaps could in theory be changed by the
parties as they please. This is uncommon.
This applies both for single names and indices, i.e. you can have dividend swaps and
dividend futures for both indices and single names. The trend over the last years has
DIVIDEND
13
No Exam
Dividend futures are equivalent to OTC dividend swaps at expiry and so the price is
driven by the same mechanics as the underlying dividend swaps.
SWAPS
DIVIDEND
14
Index Dividend Futures: Notional & Payout
“Realised” and “implied” dividends are expressed in index points (ip).
For Euro STOXX 50 dividends futures the value per 1 index point is 100€.
15
Index Dividend Futures: Realised dividends in Index Points
Index dividends are calculated by aggregating all dividends paid by the index
constituents, expressed in index points. The calculation of index dividend points is
based upon how much the index should theoretically fall after each aggregate single
stock dividend goes ex-dividend.
N
NOSH ⋅ FF i ,t ⋅ Pi ,t
∑i =1
i ,t
Divisor t
P is the stock price for company i at time t(converted to the index currency).
Divisor is the index divisor at time t. The divisor is a number that is adjusted to
DIVIDEND
Divisor t
D is the dividend per share paid by company i at time t (converted to the index
currency).
The payout for the index dividend futures is calculated as the sum of the realised dividend points
for all the days specified in the dividend futures contract
For example, for a 2015 contract this means all business days between 19-Dec-15 and 18-Dec-15
Dec 15 N
NOSH ⋅ FF i ,t ⋅ D i ,t
∑ ∑
t = Dec 14 i =1
i ,t
Divisor t
SWAPS
DIVIDEND
17
Index Dividend Futures: Realised dividends in Index Points
Corporate actions can lead to share changes and can therefore impact a company’s contribution in
index points.
Index reconstitution risk: the constituents of an index change in time. Sometimes this may be a result
of a take-over, but it can also be as a result of a periodic index review. There are two main ways that
index reconstitution can affect index dividend swaps:
Replacement stocks may have a different dividend yield or weight than the original
constituent stock, thus affecting the total index dividends paid. In practice, this risk is
typically relatively contained as the stocks that are most normally affected by replacement
tend to have a small weight. For indices with a larger number of members, such as the S&P and
FTSE, the risk from replacement is smaller than for more concentrated indices.
Dividend ex date timing risk: a stock may be replaced before it pays a dividend with a stock
that has already paid its dividend for that year. In this case, total realised index dividends will
be reduced. When Royal Dutch / Shell first announced their unification, the timing of the
merger was crucial to the 2005 Euro Stoxx realised dividends, due to the proximity of the main
European dividend season. Another example was the removal of relatively high-yielding
Telecom Italia Mobiles and Royal Dutch from the Euro Stoxx 50 in 2006, which had a negative
impact of 3.5 index points on total index dividends in that year.
SWAPS
DIVIDEND
18
Index Dividend Futures: Realised Dividend Indices
STOXX computes daily the Euro STOXX 50 realised dividend index, called Euro STOXX
50 Dividend Points (DVP).
The index resets to zero following the end of each reference period (3rd
Friday of Dec for the Euro STOXX 50). At futures expiration, dividend futures
settle at the underlying index level.
This index is also used for Euro STOXX 50 dividend options, which are also
listed and traded on Eurex.
19
Index Dividend Futures: Realised Dividend Indices
180 300
Euro STOXX 50 DVP FTSE 100 DVP - right axis
160
250
140
120 200
100
150
80
60 100
40
50
20
0 0
Jan-05
Jan-06
Jan-07
Jan-08
Jan-09
Jan-10
Jan-11
Jan-12
Jan-13
Jan-14
Jan-15
Jan-16
SWAPS
As of March 2016
FT = S ⋅ e rT − E [ Div ]
Where S is the stock price, F is the future price with expiry T, and Div is the
sum of all dividends that will go ex from today to time T, valued at time T.
References: J.P. Morgan, “Dividend Swaps”, 2009 - Section 5.1 (Put-call parity &
forwards), 5.2 (Cash & carry), and 5.3 (Forwards and dividend swaps).
DIVIDEND
22
Put-Call Parity & Dividends (I)
When pricing options, traders must therefore estimate the expected dividends that
will be paid over the life of the option. When a company pays a dividend, the share
price is reduced by the amount of the dividend, but the option’s strike price remains
the same. In order to price the option correctly, an accurate estimation of future
dividends is essential.
S − E [ Div ] ⋅ e − rT + P = C + K ⋅ e − rT
Where S is the stock price, C and P call and put prices with expiry T and strike
K, and Div are the dividends to be paid from today to time T (valued at time T).
So if we can observe the price of puts and calls in the market we can derive an
‘implied dividend’
23
Put-Call Parity & Dividends (II)
Example:
On 14-Mar-08: Euro Stoxx 50 was trading at 3566. December 3600 calls and puts
were trading at 290.6 and 349.4 respectively. Risk free rate to December was
4.6%.
How would you calculate the implied dividends corresponding to the Euro STOXX 50
Dec-17 dividend futures?
Hint: you need both Euro STOXX 50 Dec-16 and Dec-17 puts and calls.
SWAPS
DIVIDEND
24
Dividend Swap Market: Origins and evolution (I)
Dividend swaps were first traded in the late 1990's/early 2000's as a way of reducing
dividend exposure that had accumulated on structured equity derivatives providers’
books.
The issuance of “retail” structured products leaves banks with exposure to long
dated forwards that need to be hedged in order to continue issuing more
products. Dividend are a large component of risk in a long-dated equity fwd .
Hedge funds are generally the first investors to take the risks that banks need
to offload (obviously, at very attractive levels, given banks’ needs).
— Initially the product is traded OTC.
Institutional investors follow hedge funds in the use of the product, generally
with a more “fundamental” rationale.
SWAPS
25
Dividend Swap Market: Origins and evolution (II)
Which retail structured products are these?
Equity-linked products.
“Linked” meaning its performance is linked to the capital appreciation
of an equity index (or a basket of stocks)
Reference indices are often Price Return (i.e. exclude dividends)
What is the investor “really” buying? A risk-free zero coupon bond plus an
SWAPS
What is the investor “really” buying? A risk-free bond plus selling ATM puts
(often with knock-in features).
Reverse convertibles are usually created synthetically and are popular with
investors that are not able to sell options outright and expect the price not
SWAPS
Capital guaranteed notes & reverse convertibles are popular products among
investors during “bull” years.
Banks end up being short calls & long puts … (using put-call parity with divs.)
Short stocks
Long (realised) dividends
How do you hedge this exposure?
SWAPS
References: J.P. Morgan, “Dividend Swaps”, 2009 (Section 5); J.P. Morgan, “Global
DIVIDEND
28
Dividend Swap Market: Origins and evolution (V)
For a dividend-paying stock, it can be shown that (put-call parity):
S − E [ Div ] ⋅ e − rT + P = C + K ⋅ e − rT
Where S is the stock price, C and P call and put prices with expiry T and strike
K, and Div are the dividends to be paid from today to time T (valued at time T).
See Hull (6th edition, Chp. 9.7 – Effect of dividends).
If you are short calls and long puts, your exposure is given by:
P − C = − S + E [ Div ] ⋅ e − rT + K ⋅ e − rT
you are short the stock and long the realised dividends.
SWAPS
DIVIDEND
29
Dividend Swap Market: Origins and evolution (VI)
“… banks are left with hedging needs they have to take care off…”
The usage of dividend swaps increases, institutional investors get involved, the
SWAPS
30
Dividend Swap Market: Origins and evolution (VII)
Imagine you are a bank who has sold capital guaranteed notes and reverse
convertibles (e.g. one of each on the same index or stock).
Imagine the options on each of those two products are ATM options; today is
time 0 and both products expire at time T.
If you just buy the stock (one share) in order to hedge your exposure above, you will
receive realised dividends, which you don’t really need to pass on to your structured
product clients.
If, at expiry, ST > S0, you will give ST - S0 to the holder of the cap. guar. note.
If, at expiry, ST < S0, you will receive S0 – ST from the holder of the reverse
convertible
What do you do with the dividends the stock has paid from 0 to T?
SWAPS
Thus, to hedge your exposure you have to buy the stock and sell realised dividends.
31
What drives dividends? (I)
Technical factors (in Europe): Structured products
The end investors’ long equity market exposure, though, is actually through
forwards (i.e. price change only, excluding dividends) and not through spot
(i.e. Including dividends), since the investor only receives the equity-linked
return at some future date (i.e. at maturity of the structured product).
Investment banks are left with short forward positions which are basically
equivalent to a short spot and long dividend exposure.
SWAPS
Hedging of this position: buy spot and sell the future dividend stream (i.e.
sell realised dividends).
DIVIDEND
32
What drives dividends? (II)
Technical factors (in Europe): Structured products
Hedging of this position: buy spot and sell the future dividend stream.
How?
The simplest way is to buy equity forwards/futures, but equity forwards
contain exposure to interest rates, which can obscure the sensitivity to
dividends.
Selling dividend swaps.
Thus, as equity markets rally and retail investors pile into structured
products, banks recycle their dividend exposures via the dividend swap
market, which puts downward pressure on implied dividends.
Hence this dynamic between interim flow pressure and final valuation can
generate attractive investment opportunities for investors willing to warehouse
DIVIDEND
33
What drives dividends? (III)
Technical factors: Structured products – Europe & Asia vs. US
In Europe and Asia, market makers tend to be long dividends. The long-dated
option market in the US is very distinct from the long-dated option markets in
Europe and Asia. In these regions, the chief source of demand for long-dated
option exposure is the structured product market, where retail investors
effectively buy long-dated calls (through capital guaranteed access products) or
SWAPS
sell puts (through reverse convertibles). These positions leave the providers of
structured products long dividends on the underlying equities or indices,
DIVIDEND
34
What drives dividends? (IV)
The distribution of open interest of dividend
futures also is indicative that structured
products have influence over the market. We
can see this, by considering how dividend
exposures are built up by structured products
over time.
Each long dated structured product,
schematically shown below, builds up exposure
to implied dividends that is fairly linearly
distributed through the maturity of that
product. Hence as the product matures, some of
its implied dividend exposure is lost, but that
loss of exposure is be topped up by new
structured products issued. Hence at any
current point in time there should be a ladder of
future dividend exposure, with the majority of
SWAPS
35
What drives dividends? (V)
Realised dividend payments made by companies are ultimately driven by two
fundamental factors.
Ability: whether the company has generated any cash to pay a dividend.
Willingness: whether the company wishes to pay any of the cash back to
shareholders.
Since dividend swaps for single stocks are mainly limited to fairly short-dated
maturities, they should be analysed in the context of the fundamental drivers
that are likely to be specific to each stock.
36
The Pull-to-Realised effect
Much of the attraction of using dividend swaps relies on the pull-to-realised effect.
Pull-to-realised is due to the fact that dividend swaps eventually pay out
according to the actual realised dividends that companies pay.
37
Sources of Dividend Information
Dividend swap/futures market.
Futures market.
Dividend have a tangible valuation framework and can be analysed directly from a
fundamental economic perspective or using standard equity valuation models.
38
The dividend curve – long-dated and short-dated dividends
The front end of the dividend curve (up to 3 years) benefits from high visibility of
the expected dividend payouts from a variety of sources including analyst estimates,
company guidance, historical dividend trends, competitor read-across, etc.
Moreover, the dividend futures are pulled to realised as expiry approaches. The
volatility of current year dividend futures diminishes as the expiry approaches, which
is similar to the volatility dynamics of a corporate bond.
From a cash flow perspective, being long
Div vs. Bond Volatility
short-dated dividend futures could be
compared to owning a zero coupon bond,
since there is no interim cash flow until
maturity.
Rather than being comparable to bonds,
the dynamics of the longer dated
dividends are much more equity-like.
The visibility on long dated dividends is
SWAPS
Noticeably, at that
time, the share price
was unchanged after
the announcement
and so a positive view
of the dividend was
best (only) captured
through taking the
dividend swap
SWAPS
exposure.
DIVIDEND
Div Future 2016 Div Future 2017 Div Future 2019 Euro STOXX 50
5%
0%
-5%
-10%
-15%
-20%
-25%
Feb-16
Mar-16
Jun-15
Jul-15
Aug-15
Sep-15
Oct-15
Nov-15
Dec-15
Jan-16
SWAPS
Euro STOXX 50 div upside (29-Mar-16) Expected upside on dividends (since 06)
Implied Dividends vs. Estimates and Potential Upside
Potential Potential Potential 4,800
Source 2016 Upside 2017 Upside 2018 Upside
IBES bottom up estimates for next year divs %Over/(Under) estimate of IBES as % of
(index points) realised dividend
200
40%
2009
2008 30%
2009
20% 2010 2012
150 2007 2013
2010 2012
2011 10% 2008
2006 20132014 2015 2007 2011
0% 2006
100 2005 -10% 2005
-20%
-30%
50
Jan 04
Jan 05
Jan 06
Jan 07
Jan 08
Jan 09
Jan 10
Jan 11
Jan 12
'04
'05
'06
'07
'08
'09
'10
'11
'12
'13
'14
SWAPS
DIVIDEND
45
DOCUMENTATIONAL AND LEGAL CONSIDERATIONS
Agenda
Page
CDS as a Contract 1
Case Studies 13
1
What is a Credit Default Swap?
CDS as a Contract
Originally, the contracts were bespoke and drafted for particular situations.
Today we have a standardised contract that allows market participants to originate and offset
trades easily.
2
Credit Events
Credit Events
Credit events:
Bankruptcy – covers bankruptcy filings, insolvency, appointment of administrators/liquidators
for the reference entity.
Failure to pay – when the reference entity misses a coupon or principal payment on any
obligation, taking a grace period into account.
Restructuring – a change to specific bond terms as a result of a credit deterioration. Includes
reduction of principle, reduction of coupon, change to maturity, currency or seniority.
– North American CDS contracts do not have restructuring as a separate credit event;
companies usually enter Chapter 11 bankruptcy instead.
Governmental Intervention (GI) – a change to a bond term imposed by a governmental
agency.
– Applicable to Financial CDS.
Western European Sovereigns have different credit events: Failure To Pay, Restructuring,
CDS AS A CONTRACT
Repudiation/Moratorium.
3
Obligations
Obligations
Obligations
Credit Events are only triggered by actions on defined Obligations
Obligations typically cover “borrowed money” – e.g. bonds, loans, deposits.
– An equity or pref share is not an obligation.
Trade payables are not borrowed money.
Non-standard CDS contracts can have different definitions of obligations – e.g. a single bond.
Deliverable Obligations
Bonds and Loans are deliverable – these must be issued or guaranteed by the Reference Entity
The Reference Obligation is deliverable
– A specific bond referenced in the confirmation.
– as such it doesn’t need to meet the general deliverable criteria.
Deliverable Criteria – the contract sets out rules about the Bonds and Loans that can be
delivered dealing with: seniority, currency, contingent obligations, conditionality of guarantees,
transferability etc.
CDS AS A CONTRACT
For example, a Thai Bhat bond is not deliverable unless the contract explicitly allows it, or it
is the Reference Obligation.
4
Succession Events
Percentages calculated by reference to Bonds and Loans outstanding prior to the announcement
of the transaction
Very important term of the contract for dealers, investors and issuers to understand because of the
economic consequences.
New Co. A
Original
Reference
Entity ?
CDS AS A CONTRACT
New Co. B
Source: J.P. Morgan
5
Effective Date
The effective date (i.e. date from which the protection starts) for all CDS contracts is a rolling x-60
days for Credit Events and x-90 days for Succession Events.
Contracts are fungible: All contracts have identical effective dates on any given day.
You can think of it like a tail that moves as we move forward in time.
Contract 1
Today – 60
calendar days
Contract 2
Today – 60
calendar days
Due to the dynamic look back period the DC doesn’t need to make the determination by the
60th/90th day but it does need to have been notified of the event by that deadline.
CDS AS A CONTRACT
6
The ISDA Determinations Committee
The ISDA DC
Some credit events are straightforward (e.g. bankruptcy filings) but some can be much more
subtle.
For standardised CDS contracts, credit events are decided upon by a group called the ISDA
Determinations Committee.
Set up in 2009 as part of the “Big Bang” changes to CDS documentation.
15 members – 10 sell-side and 5 buy-side.
Need a supermajority of 80% to agree on a credit event for trigger.
If 80% majority cannot be reached, case goes to three member arbitration panel.
Vast majority of historical DC decisions have been unanimous.
7
Agenda
Page
CDS as a Contract 1
Case Studies 13
8
Restructuring
1. A change to the money terms of the obligation that is not allowed under the existing
documentation.
Reduction in Interest or principal
Change of Currency
Maturity Extension
Subordination
RESTRUCTURING AND GOVERNMENTAL INTERVENTION
3. Credit Deterioration
Low bar, but needs to be shown.
9
Restructuring
The pool of deliverable bonds is reduced to those less than your CDS maturity.
In the US, Restructuring is not a Credit Event for standard contracts. Restructuring generally
falls under Chapter 11 which is covered by Bankruptcy.
In Sovereign CDS there is no maturity limitation.
The provisions only apply if the Credit Event Notice is delivered by the Buyer.
i.e if the seller triggers the contract, then any debt up to 30-years can be delivered.
10
Governmental Intervention
Governmental Intervention is a new bail-in CDS trigger to be introduced for Financial (Bank) CDS
contracts from September 2014.
It has arisen because bank bonds have changed in recent years to allow write-down by a regulator
or governmental authority under certain conditions.
Investors want protection against these events.
1. Bonds may have explicit write down language in the bond documentation.
• If a bond allows a change to the terms, then it is not a Restructuring Credit Event.
• This would mean that there are no obligations left to deliver into the contract.
11
Governmental Intervention
Transaction Type
Sub CDS contracts will no longer be senior contracts with sub deliverables but will be
differentiated by transaction type with "Senior Transactions” and "Subordinated Transactions”.
Credit Events - "Senior Transactions” will only trigger if senior bonds are impaired.
Successor Provisions - CDS will track its debt tier in a succession event with senior CDS
tracking senior bonds and sub CDS tracking subordinated bonds.
Source: J.P. Morgan
12
Agenda
Page
CDS as a Contract 1
Case Studies 13
13
Credit Events Case Study
Lehman 2008
Lehman Brothers filed for Chapter 11 Bankruptcy on Monday 15th September 2008.
Greece 2012
In 2011 Greece offered a voluntary exchange to its bondholders; investors could voluntarily swap
accept a 50% haircut to their bond notional and instead receive a package of new Greek bonds,
AAA Eurozone-backed bonds and cash.
Despite this large haircut, this bond exchange did not trigger CDS as it was voluntary in nature.
This bond exchange did not achieve sufficient take-up from investors – in March 2012 Greece
forced all bondholders to accept the exchange through an act of law, triggering a restructuring
credit event.
14
Agenda
Page
CDS as a Contract 1
Case Studies 13
15
Settling CDS Trades
CDS Settlement
Following a credit event, the protection buyer will receive a payout from the protection seller.
This payout is equal to (1 – bond recovery) x Notional.
How do we determine what recovery is?
16
The CDS Auction
Physical Settlement.
Mechanic allows Buyer to exit a position in the security and Seller to enter a position in the security.
Cash Settlement.
17
Examples
18
Auction Pricing is a Two-Step Process
Step 1 Step 2
A) Each dealer submits a bid and offer market Based on the net open interests, Dealers submit
(Valid Inside Market Submission) Limit Orders from themselves and clients to
buy/sell bonds
B) Each dealer submits a notional of bonds to
physically buy/sell (sum of Dealer Physical If the net Open Interest was to sell bonds, then
Settlement Request and Client Physical Settlement the lowest bid from the Limit Orders is the Final
Request) Price for everyone
Administrator calculates If the net Open Interest was to buy bonds, then
the highest offer from the Limit Orders is the
a) initial recovery rate (Inside Market
Final Price for everyone
Midpoint) (from A) and
19
Auction Pricing is a Two-Step Process – Step 1A
Each dealer submits a bid and offer market Administrator sorts bids from highest to lowest
(Valid Inside Market Submission) and offers sorted from lowest to highest.
Tradeable markets identified and discarded from
the Inside Market Midpoint calculation
A 39.5 41 D 45 34 E
Tradeable markets –
B 40 42 C 41 39.5 G
these Dealers are
C 41 43 H 41 40 F penalised
D 45 47 B 40 41 A
Best half of remaining
SETTLEMENT AND CDS AUCTIONS
E 32 34 A 39.5 42 B
prices used to determine
F 38.75 40 F 38.75 42.75 H
Inside Market Midpoint
G 38 39.5 G 38 43 C
H 41 42.75 E 32 47 D
IMM = 40.625
20
Auction Pricing is a Two-Step Process – Step 1B
Administrator calculates the Net Open Interest In this case there are €500m of Sell Physical
based on the Physical Settlement Sell and Buy Requests and €150m of Buy Physical Requests
requests
Net Open Interest is therefore €350m to sell
bonds
A Offer 150 Dealers D,C and H are penalised for having bid
B Offer 200 too high
C Offer 10 €500m Sell
Physical
D Offer 40
Requests
E Offer 50
F Offer 50
SETTLEMENT AND CDS AUCTIONS
€150m Buy
G Bid 20
Physical
H Bid 130
Requests
21
Auction Pricing is a Two-Step Process – Step 2
Dealers submit Limit orders from themselves The Limit Orders are matched against the Open
and clients Interests
E 39.875 30
Matched Limit Orders
G 39.625 20
SETTLEMENT AND CDS AUCTIONS
F 39.375 50
C 39.125 30
22
Settlement for Restructuring
Restructuring
Triggering a restructuring contract is optional, neither party may wish to trigger the contract.
Since deliverable obligations are limited to the CDS contract maturity, you could get numerous
different recovery rates.
Solution is to form buckets around maturities and a different auction is held for each maturity
bucket.
Each maturity bucket limits the maturity of the deliverable obligations that can be delivered for
that bucket.
23
Agenda
Page
CDS as a Contract 1
Case Studies 13
Online Resources 24
24
CDS Online Resources
CDS resources
www.markit.com/cds
CDS data.
Index data and rules
www.dtcc.com/products/derivserv/suite/ps_index.php
Data on Credit Derivatives outstanding volumes & trading activity.
www.isda.org/credit
Requests to the Determinations Committee, decisions and information about credit events.
SETTLEMENT AND CDS AUCTIONS
www.creditfixings.com
Information about all the CDS auctions that have been held by ISDA.
25
STRICTLY PRIVATE AND CONFIDENTIAL
Page
Recovery Products 1
Quanto CDS 19
1
Fixed Recovery CDS
Conventional CDS contracts pay out par minus recovery upon a credit event, where the recovery is determined by a auction
process.
The eventual recovery is not known in advance – the value of a CDS can differ greatly depending on the expectations of the
recovery rate.
Fixed Recovery CDS is a special type of CDS contract where the recovery rate used to determine the CDS payout is fixed and
agreed upon by the trade counterparties at the inception of the trade.
Example:
We buy fixed recovery CDS protection on ABC Corp at 100bp with a fixed recovery of 20%.
ABC Corp later has a credit event and the auction gives a recovery rate of 45%.
The fixed recovery contract may trade with a different spread to the conventional contract.
For example, a zero recovery contract is likely to trade wider than a normal contract as the default payoff will be equal to the
full notional of the contract.
RECOVERY PRODUCTS
2
Pricing Fixed Recovery CDS – Full Running Spread Convention
The original CDS pricing equation gives us the present value of the contract for a buyer of protection:
Maturity Maturity
PV = (1 − R) ∑ ( PS i −1 − PS i ) DFi − S ∑ ∆ i PS i DFi − AccrualOnDefault
i i
We can also simplify this by writing in terms of the risky annuity, which is a function of the time to maturity, survival
probabilities and interest rates RA=RA(λ,r,T). The risky annuity is not a direct function of spreads or recovery.
PV = (1 − R )λ × RA(λ , r , T ) − S × RA(λ , r , T )
We know that for a conventional CDS contract trading on full running spread, the par spread is equal to, S Par = λ (1 − R )
where R is the recovery being assumed for a convention contract.
For a fixed recovery contract trading at a spread of Sfixed and a recovery of Rfixed the pricing equation becomes:
For example, if a conventional CDS trades with a full running CDS spread of Spar = 200bp and an assumed of recovery of
40% then a fixed recovery contract with Rfixed = 0% would trade with a full running spread of Sfixed = 333bp
Source: J.P. Morgan
3
Pricing Fixed Recovery CDS – Fixed Coupon + Upfront
When calculating the upfront for a Fixed Recovery CDS trade the equation is the same as for a conventional trade:
For a risky annuity of 4.5 and coupon of 100bp, the fixed recovery trade on the previous page would have an upfront of
10.5% (=(333bp-100bp) x 4.5) whereas the conventional contract would have an upfront of 4.5% (=(200bp-100bp) x 4.5).
If we choose Rfixed to be equal to the market assumed recovery rate R then the upfront is identical to a conventional contract
and the entire fee leg will be identical.
If we buy protection on a fixed contract and sell protection on a conventional contract with the same upfront then the feel legs
cancel out and we only have exposure to the default leg.
This gives us a direct exposure to recovery values – the combined package is known as a recovery swap.
4
Recovery Swaps
The fixed recovery contract is struck at a level which gives Sell protection
it the same upfront as the standard contract. Fixed recovery CDS
Fee Leg
Only cashflow is after a credit event.
Contigent Leg
Recovery swaps have zero upfront cost and no coupons Payment on default
during life of trade. (1 - Fixed Recovery Rate)
Buy recovery: think recovery will be higher than fixed level Sell recovery Payment on default
Recovery Swap (Fixed Recovery Rate -
Buy fixed recovery CDS protection Auction Recovery Rate)
Sell standard CDS protection
Contigent Leg
Sell recovery: think recovery will be lower than fixed level Fee Leg
RECOVERY PRODUCTS
If there are no deliverable obligations, the final recovery Source: J.P. Morgan
5
Recovery Swaps – An Example
In two years time OTE has a Failure To Pay credit event and undergoes a CDS auction.
6
Marking Recovery Swaps to Market
Recovery Swaps
A recovery swap is constructed from a conventional CDS contract and a fixed recovery contract.
The recovery of the fixed recovery contract is chosen so that the fee legs of the two contracts cancel out.
The recovery swap can be marked-to-market by calculating the present value of the two contingent legs.
Present value of contingent leg for fixed recovery CDS = λ (1 − RFixed ) × RA(λ , r , T )
Present value of contingent leg for conventional CDS = λ (1 − R) × RA(λ , r , T )
where Rfixed is the agreed upon fixed recovery, R is the assumed recovery for a conventional CDS contract and λ is the
annual probability of default.
For a buyer of recovery (i.e. buy fixed recovery protection and sell conventional protection) the present value of a recovery
swap is given by:
PV = λ (1 − RFixed ) × RA(λ , r , T ) − λ (1 − R) × RA(λ , r , T )
= ( R − RFixed ) × λ × RA(λ , r , T )
This can be thought of as the difference in the expected recovery payouts, multiplied by the probability of receiving that
payout over the lifetime of the contract.
7
Recovery swap example
Recovery Swaps
We sell recovery on OTE at 35%. The initial mark-to-market (ignoring bid/offer) of this swap is zero, given that the fixed
recovery Rfixed is initially equal to the market assumed recovery R.
One month later the market’s expectation of recovery falls to 20%. The current 5y par spread is 500bp and the risky annuity is
4.0.
The mark-to-market on the recovery swap is equal to 3.75% of notional (=(35% - 20%) x 6.25% x 4).
Recovery swaps give us direct exposure to recovery levels, but also have a secondary exposure to credit spreads and default
probabilities.
As spreads and default probabilities increase, the magnitude of the mark-to-market increases whether it is positive or
negative.
Increasing the default probability effectively increases the likelihood that the current difference in fixed and assumed
recoveries will be realised.
Source: J.P. Morgan
RECOVERY PRODUCTS
8
Why trade recovery products?
Recovery products
When we buy or sell conventional CDS protection, we are effectively taking a view on both default probabilities and recovery.
If we buy protection, we benefit from rising default probabilities and/or falling recovery expectations.
Fixed recovery CDS and recovery swaps take exposure to just one of these factors:
Fixed recovery CDS only has exposure to default probabilities.
Recovery swap cashflows only have exposure to the entry and final recovery rates.
recovery swap.
This will be much cheaper than buying CDS protection; recovery swaps have no upfront costs and no spread to pay
during the life of the trade.
Source: J.P. Morgan
9
Historical recovery distributions
70
Allied Irish Irish Life
Thomson
60
50
40
30
Greece
20
Jun-08 Dec-08 Jul-09 Jan-10 Aug-10 Feb-11 Sep-11 Apr-12 Oct-12 May-13 Nov-13
10
Agenda
Page
Recovery Products 1
Quanto CDS 19
11
iBoxx Corporate Bond Indices
iBoxx
These track the total return – including price changes and coupons – of different types of corporate bonds.
For example, the iBoxx € Corporates index tracks the performance of the Euro-denominated investment grade corporate
bond market.
The quoted index number is a total return index. The percentage difference in this number over a certain period gives the
total return that would have been made from buying all the bonds in the index.
For example:
On 31st December 2012 the index level was 192.22.
On 31st December 2013 the index level was 196.52
The total return of the Euro-denominated investment grade bond market over 2013 was 2.24% (= 196.52/192.22 – 1).
Each index has different inclusion rules – these are available from www.markit.com
iBoxx is not the only family of corporate bond indices; other commonly-used indices are run by JPMorgan, Barclays and
Merrill Lynch.
12
iBoxx Total Return Swaps
iBoxx Total Return Swaps (TRS) allow investors to take macro exposure to corporate bonds in a single contract.
TRS on corporate bond indices have been traded on and off for the last 10 years but only really took off in 2011.
3mEuribor
IBOXX TOTAL RETURN SWAPS
13
iBoxx Total Return Swaps
We buy the index return through a 3-month TRS on the iBoxx € Liquid HY index maturity on the 20th June 2014 on a notional
IBOXX TOTAL RETURN SWAPS
of €50million.
We enter the trade at an index level of 160. The 3m Euribor fixing is equal to 0.25%.
We receive an index leg payout of €937,500 (=163/160 x €50mm) and pay a floating leg of €31,250 (=0.25% x ¼ x €50mm).
14
What determines the entry and exit levels?
To calculate the index leg payout we need to know the entry and final levels of the relevant iBoxx index.
The final level is simply the total return index level published by Markit at the close of business on the day of the maturity of
the swap.
The entry level is agreed by the index buyer and seller at trade entry.
This is usually close to the level published by Markit on the entry date, but demand and supply can cause it to trade above
or below the Markit-published level (often referred to as the Net Asset Value or NAV).
For example, there is currently a large client demand to invest in European high yield through TRS.
This has driven the quoted entry levels on iBoxx Total Return Swaps above the Markit-published NAV.
premium.
15
Uses of iBoxx TRS
Real Money accounts (e.g. asset managers, pension funds, insurance companies)
iBoxx TRS is a much better replication of their portfolios than CDS indices.
Common uses include:
– Inflow management
– Asset allocation
– Portfolio hedging
16
Trading iBoxx TRS vs. comparables
* This is not an exhaustive comparison between the various credit strategies. The information is provided for illustrative purposes only and investors should account for various risks between the strategies, including but not limited to counterparty,
margin, and liquidity considerations
ETFs can have a large tracking error to the underlying index… … and the bond-CDS basis means that CDS is not a perfect hedge
4%
2%
0%
-2%
IBOXX TOTAL RETURN SWAPS
-4%
-6%
-8%
Sep-10 Dec-10 Mar-11 Jun-11 Sep-11 Dec-11 Mar-12 Jun-12 Sep-12 Dec-12
Source: J.P. Morgan Source: J.P. Morgan. P&L shown for 10 HY corporate bond portfolios hedged with iBoxx TRS
and iTraxx Xover.
17
iBoxx and iTraxx Tracking Errors
iBoxx TRS has very little tracking error to bond portfolios iTraxx Main is a much less accurate hedge for bond portfolios
70% 20%
60%
50%
40% 0%
30%
20%
10% -20%
0%
-10%
-20% -40%
-30% Dec-07 Dec-08 Dec-09 Dec-10 Dec-11 Dec-12 Dec-13
Dec-07 Dec-08 Dec-09 Dec-10 Dec-11 Dec-12 Dec-13 Hedging with iTraxx + £ Swaps Hedging with iBoxx TRS
Average Bond Portfolio iBoxx £ Corporates iTraxx + £ Swaps
Source: J.P. Morgan. Average return of 10 £ IG Corporate Bond Funds run by SWIP, Standard Source: J.P. Morgan. Based on Average return of 10 £ IG Corporate Bond Funds run by by
Life, Invesco, Alliance Trust, Fidelity, Aviva, Henderson, M&G, L&G and Insight. SWIP, Standard Life, Invesco, Alliance Trust, Fidelity, Aviva, Henderson, M&G, L&G and
Insight.
Comparison of Fund Tracking Error and Correlation with iBoxx € Liquid High Yield and iTraxx Crossover
18
Agenda
Page
Recovery Products 1
Quanto CDS 19
19
Quanto CDS
Quanto CDS
We have so far concentrated on CDS denominated in the home currency of the issuer.
This is most relevant for sovereigns: Western European sovereigns typically trade in both EUR and USD.
The difference in spread between two different-currency contracts is known as the quanto.
If we want to buy protection on a notional of €10million we could also buy this on an equivalent notional of USD-denominated
CDS.
We may prefer to buy one contract over the other – this is generally the result of a correlation between the credit spread and
the foreign exchange rate.
If we were going to buy protection on Germany, would you rather buy a USD or EUR contract?
What do we expect to happen to the Euro if Germany defaulted?
We would expect the Euro to greatly depreciate if Germany defaulted, so we would rather buy protection on the USD
contract. As a result, we would expect the USD contract to trade at a wider spread.
QUANTO CDS
20
Quanto CDS
USD-denominated contracts generally trade with a wider Country USD EUR Quanto (bp) Quanto (%)
spread than EUR-denominated contracts for names whose Italy 113 86.5 -26.5 -31%
credit equality is expected to decline as the Euro Spain 89 57 -32 -56%
depreciates. Ireland 73 54.5 -18.5 -34%
Portugal 170 145.5 -24.5 -17%
The higher the correlation, the bigger the quanto. Germany 21 10.5 -10.5 -100%
UK 21.5 16 -5.5 -34%
Eurozone core nations generally have a higher quanto than France 46.75 28.25 -18.5 -65%
the peripheral countries as a default in a core nation is Austria 36.5 22 -14.5 -66%
expected to have a larger systemic impact on the value of Netherlands 30.5 18.5 -12 -65%
Belgium 40.5 23 -17.5 -76%
the Euro.
Source: J.P. Morgan
How can we quantify the quanto in terms of the correlation
between spreads and FX?
21
Quanto CDS Pricing
Quanto CDS
To model the size of the quanto spread we put ourselves in the position of a Euro investor buying USD-denominated CDS
protection.
We need to calculate the expected loss of the USD denominated contract; comparing this with the expected loss of the EUR
contract will allow us to calculate the size of the quanto.
There are two separate factors that can contribute to the quanto:
Depreciation on default
If the reference entity suddenly defaults overnight how much depreciation can be expected in the currency?
Slippage
As a EUR investor, if we buy USD protection and want to maintain a constant exposure in EUR we will have to rescale the USD
notional of the trade by buying/selling additional protection as the FX rate changes.
If the EUR depreciates, we need to sell protection. If the EUR appreciates, we need to buy protection.
If there is a correlation between spread widening and the EUR depreciating, we are likely to be selling protection when spreads
are high and buying protection when spreads are low. Both of these are beneficial – if the EURUSD rate fluctuates in a range
we will make money consistently from buying low and selling high.
This represents a positive convexity premium in our favour; the seller of protection will demand that we pay them a higher
spread in order to compensate them for this.
Expected Loss (USD contract) – Expected Loss (EUR contract) = Expected Loss due to Depreciation on Default + Slippage
Costs
QUANTO CDS
22
Depreciation on Default
Quanto CDS
The additional expected loss due to depreciation on default can be written as the PV of the contingent leg multiplied by the
expected depreciation on default.
We define FXDoD as the expected depreciation of the standard currency against the non-standard currency in percentage
terms upon a credit event.
For example, if we expect the euro to depreciate against the dollar by 5% on default, then the USD-denominated contingent leg
payout is worth 5% more to us then the EUR payout would be.
= FXDoD × S × RA
i.e. the additional expected loss due to the depreciation on default is equal to the expected depreciation in the standard
currency against the non-standard currency multiplied by the par spread of the non-standard currency contract multiplied by the
risky annuity.
For example, if the expected depreciation on default is 5%, the par spread is 200bp and the risky annuity is 4.5, then the
depreciation on default contributes an additional 0.45% to the expected loss of the non-standard currency contract.
23
Slippage
Slippage Costs
The P&L due to a change in spread on a USD-denominated CDS for a Euro-based buyer of protection is given by:
P & L = ∆S N × RA × FX 0 / FX 1
where ΔS is the change in spread of the non-standard currency, RA is the risky annuity, FX0 is the initial EURUSD exchange
rate and FX1 is the EURUSD exchange rate after the spread move. We assume that the daily spread changes on the two
contracts are the same, such that ΔSN = ΔS.
where ∆
%, d
FX is the day’s percentage change in the FX rate.
Using this, we can write the expected value of the day’s P&L due to spread changes as:
24
Slippage
Slippage Costs II
We can rewrite the covariance as Cov ( X , Y ) = ρ XY σ X σ Y , where ρ is the correlation between the two variables and σ is
the volatility.
We can rewrite the daily basis point spread volatility and daily percentage FX volatility in terms of the annualised percentage
volatility as follows:
25
Slippage
To take discount factors and survival probabilities into account we approximate the sum of expected risky annuities as the
following:
RA2
∑i Ε(RA) ≈ 252 ×
2
This gives us the following expression for the total additional expected loss due to slippage:
RA2
= − ρS Nσ S σ FX
%, y %, y
2
If the correlation is negative (i.e. if the standard currency depreciates as spreads widen) then this will result in the non-standard
currency trading with a higher expected loss than the standard contract.
Combining this with the expected loss due to the depreciation on default, the difference in expected loss between a non-
standard and standard currency contract can be written as:
RA2
= FXDoD × S × RA − ρSσ S σ FX
Expected Loss (Non-Standard) – Expected Loss (Standard)
%, y %, y
2 2
%, y RA
(S N − S ) × RA = FXDoD × S × RA − ρSσ S σ FX
%, y
2
The quanto as a percentage of spread is therefore approximately equal to:
SN − S RA
Quanto(%) = = FXDoD − ρσ S%, yσ FX%, y
S 2
For example, if the expected EURUSD depreciation on default is 20%, the correlation between the spread and EURUSD is
-50%, the spread volatility is 40%, the FX volatility is 15% and the risky annuity is 4.5, we would expect the USD CDS contract
QUANTO CDS
26
The Historical Quanto
27
Agenda
Page
Recovery Products 1
Quanto CDS 19
28
Additional Material
Additional Material
Recovery Products:
Recovery Swaps: Trading and hedging corporate bond recovery, J.P. Morgan, 2013
www.creditfixings.com – Historical data on CDS auction recoveries
Quanto CDS
Trading credit in different currencies via Quanto CDS, J.P. Morgan, 2010.
29
March 2016
Davide SilvestriniAC
davide.silvestrini@jpmorgan.com
+44(0) 20 7134 4082
This presentation was prepared exclusively for instructional purposes only, it is for your information only. It
is not intended as investment research. Please refer to disclaimers at back of presentation.
Equity Variance Swaps
Trading volatility
Trading volatility via straddles and delta-hedged options: path dependent P&L
Variance Swaps: Mechanics, P&L, vega notional, MtM, caps, pricing, variance swap
indices (VIX, VSTOXX, VDAX)
Volatility Swaps
Relative value
SWAPS
1
Realised Volatility: Definition and Characteristics (I)
We define volatility as the annualised standard deviation of the (log) daily return
of a stock (or index) price, and variance as the square of the standard-deviation
2
252 T Si
σ =2
T
∑ ln σ = σ2
i =1 S i −1
We compute the standard deviation over a fixed period of time (T days) and
then annualise it by multiplying it by the square root of the number of trading
days in a year (252) divided by the number of days in the calculation period.
We assume that the mean of the log daily return is zero in order to simplify
calculations (and because this is the measure used in the payoff of variance and
volatility swap contracts).
SWAPS
VARIANCE
2
Realised Volatility: Definition and Characteristics (II)
Standard deviation and variance
Moreover, when trading “vol” via delta-hedged options, the P&L is a direct
function of the difference between realised and implied variance.
Variance swaps’ payoffs are defined in terms of realised variance. However, the
market standard is to use volatility when providing a price.
SWAPS
VARIANCE
3
Realised Volatility: Definition and Characteristics (III)
Principal characteristics of volatility:
50%
40%
30%
SWAPS
20%
10%
VARIANCE
0%
1928 1938 1948 1958 1968 1978 1988 1998 2008
70%
Euro STOXX 50 3M ATM implied vol
60% Euro STOXX 50 3M realised vol
50%
40%
30%
20%
10%
0%
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
SWAPS
Plain vanilla options: gain liquidity after Black & Scholes’ (BS) option pricing
framework (1974).
Second generation:
Variance and volatility swaps emerge in the late ‘90s. Seminal papers: 1998
Carr & Madan (Towards a theory of volatility trading), 1999 Derman et al.
(More than you ever wanted to know about volatility swaps).
Third generation:
6
How can we trade volatility?
We want to make money if realised volatility (or variance) within a future time
period is higher than a given amount.
7
Variance Swaps: How we will introduce them
Unlike vanilla options, variance swaps are said to provide “pure” exposure to
volatility, in the sense that their P&L at expiry is only a function of realised volatility:
If you buy a variance swap with notional N and expiry T, your payoff at T will be
equal to N times the difference of realised volatility up to T and a fixed (pre-
agreed) volatility strike.
In order to highlight the differences between vanilla options and variance swaps we
will first illustrate the traditional alternatives to take volatility exposure via options.
Our objective is to find a “way” to obtain, via options, a volatility exposure similar to
the one provided by a variance swap.
Apart from being a useful way of introducing the rationale behind variance
swaps, this will illustrate how we can replicate a variance swap via vanilla
options.
SWAPS
This replication strategy is the backbone of variance swaps hedging (by dealers)
VARIANCE
as well as pricing.
8
Trading volatility via straddles and delta-hedged options
SWAPS
VARIANCE
9
Long ATM Straddle (I)
Buy an ATM call and ATM put. Using BS, its cost depends on: implied vol and time to
expiry (ignoring rates).
Implied volatility exposure: If implied vol increases, other things equal, the position
makes money. However, if the position is kept until expiry, the payoff is independent
of implied vol movements.
What is the exposure of this position to the realised volatility until expiry?
10
0 -50%
-10
VARIANCE
-20 -100%
50 70 90 110 130 150 50 70 90 110 130 150
X-axis: stock price.
Source: J.P. Morgan. 10
Long ATM Straddle (II)
What is the exposure of this position to the realised volatility until expiry?
Imagine realised volatility is very large, but the stock price at expiry is equal to
the strike of the straddle.
We lose money.
This isn’t what we wanted.
Initially, the delta of our position is zero, but once the stock moves away from
the strike price, the delta is not zero anymore and we have exposure to the
underlying price of the stock.
SWAPS
11
Long delta hedged option (I)
Let’s analyse the P&L of buying an option and delta hedging it during a small time
interval (e.g. 1 day) first
In order to compute the delta of the option we need to rely on a pricing model.
BS is the most commonly used
Notation:
For simplicity, we
Option price Ct assume interest rate
Stock price St and implied
volatility are
Interest rate r (= 0) constant. This
σi
allows us to ignore
Implied volatility rho and vega.
Delta ∆t Moreover, we
Γt assume interest
SWAPS
Vega Vt
12
Long delta hedged option (II)
1 day goes by ∂t (in years) and the stock price moves to S t + ∂t
We bought a call option and sold ∆t units of the underlying stock
P & Lt = [C t + ∂t − C t ] − ∆ t ⋅ [S t + ∂t − S t ]
The option price depends on the stock price, time to expiry and implied volatility. We
use an approximation for the option price change (Taylor expansion with respect to
the stock price, time and volatility ) – i.e. we assume that the option price change is
driven exclusively by Delta, Gamma, Theta and Vega sensitivities:
∆ t ⋅ (S t + ∂t − S t ) + 2 ⋅ Γt ⋅ (S t + ∂t − S t ) + Θ t ⋅ ∂ t + Vt ⋅ ∂ σ i
1 2
SWAPS
− ∆ t ⋅ [S t + ∂t − S t ]
VARIANCE
13
Long delta hedged option (III)
Assuming implied volatility stays constant (∂σ i = 0) , the P&L can be approximated by
P & Lt = ⋅ Γt ⋅ (S t + ∂t − S t ) + Θ t ⋅ ∂ t
1 2
2
Under BS, there is a one-to-one relationship between theta and gamma (assuming
zero interest rates; see Hull, 6th edition, Chp. 15.7):
1
Θ t = − ⋅ Γt ⋅ S t ⋅ σ i
2 2
2
This leaves the daily P&L of a delta-hedged call option as:
S 2
t + ∂t − S t
1
SWAPS
P & Lt = ⋅ Γt ⋅ S t ⋅
2
− σ i2 ⋅ ∂ t
2 St
VARIANCE
14
S t + ∂t − S t S
≈ ln t + ∂t
Long delta hedged option (IV) St St
S 2
t + ∂t − S t
1
P & Lt = ⋅ Γt ⋅ S t ⋅
2
− σ i2 ⋅ ∂ t
2 St
Thus, buying a delta hedged option we make money if the realised variance is above
VARIANCE
the implied one. The P&L is also affected by the “dollar gamma” of the option.
15
Dollar Gamma
Using BS, we can derive a theoretical closed form solution for the dollar gamma. It
depends on:
Γt ⋅ S t2 , where
φ (d 1 )
Γt =
St ⋅σ i ⋅ T − t
ln( S t / K ) + ( r + σ i2 / 2 ) ⋅ (T − t )
d1 =
σi ⋅ T −t
where φ (⋅) is the density function of a N(0,1), K is the strike price and T is the expiry.
SWAPS
16
Gamma: Call & Put
Call and put options (same strike, expiry and implied vol) have the same gamma;
thus, the P&L of buying (and delta-hedging) a call or a put option is the same.
Call Put
60 Cost today 60 Cost today
PnL at expiry PnL at expiry
40 40
20 20
0 0
-20 -20
50 70 90 110 130 150 50 70 90 110 130 150
120% 0%
100% Delta -20% Delta
80% -40%
60% -60%
40% -80%
20% -100%
0% -120%
50 70 90 110 130 150 50 70 90 110 130 150
4% Gamma 4%
SWAPS
Gamma
3% 3%
2% 2%
VARIANCE
1% 1%
0% 0%
50 70 90 110 130 150 50 70 90 110 130 150
17
Dollar Gamma is not constant
Dollar Gamma & Stock Price Dollar Gamma & Time to Expiry
3% Gamma 250 800 1y to expiry
Dollar Gamma (RHS) 700 6m to expiry
2% 200
600 1m to expiry
2% 150 500
400
1% 100 300
1% 50 200
100
0% 0 0
50 70 90 110 130 150 50 70 90 110 130 150
X-axis: stock price. X-axis: stock price.
Strike K Strike K
The total P&L of the trade is a function of the difference between realised and
implied variance. However, this is “polluted” by the dependence of dollar gamma on
the time to expiry and stock price.
This causes the P&L to be path dependent and, as a consequence, delta hedged
SWAPS
19
Total P&L - Delta-Hedged Option (I)
Consider a long option position on a 6-month ATM call, delta-hedged everyday to
expiry. Implied volatility of the option is set at 30% and we simulate the underlying
stock price evolution based on a realised volatility of 30% (over the 6m holding
period). This simulation is repeated 1,000 times to allow for different possible
evolutions of the underlying price.
average.
VARIANCE
The contribution to the total P&L of (realised minus implied) variance on a given
day depends on the dollar gamma for that day, which is very sensitive to the time to
expiry and the stock price.
For example, if the stock price is close to the strike during the last part of the
option’s life, whatever happens during that period has a very large impact on the
total P&L.
If we had bought the option an the stock realises very low volatility during that
period (much lower than the implied), this will have a very negative impact on
SWAPS
The final P&L can be negative even if the realised volatility since inception to
VARIANCE
expiry was very large (making the total realised volatility higher than the implied
one).
21
Total P&L - Delta-Hedged Option (III)
Example: an option trader sells a 1-year call struck at 110% of the initial price on a
notional of $10,000,000 for an implied volatility of 30%, and delta-hedges his position
daily.
The realized
volatility (over
the option’s
life) is 27.50%,
yet his final
trading P&L is
down $150k.
Why?
SWAPS
VARIANCE
S 2
t + ∂t − S t
1
P&L=∑ ⋅ Γt ⋅ S t ⋅
2
− σ i ⋅ ∂ t
2
t 2 St
Notice that every single day counts:
For each t, what matters is the combination of (i) dollar gamma for that day
and (ii) difference between stock price % change (squared) and implied vol.
Although realised variance over the life of the option may be higher than realised ...
2
There will likely be many days where S t + ∂t − S t
− σ i ⋅ ∂ t is negative.
2
St
If those days coincide with a very large dollar gamma, they can have a large
impact on the final P&L.
SWAPS
S − St
2
Especially if, for the days where t + ∂t − σ i2 ⋅ ∂ t is positive, the dollar
gamma happens to be very low.
St
VARIANCE
24
Total P&L - Delta-Hedged Option (VI)
Example Nov-01/Nov-02; 1y EuroStoxx options.
Index was initially at 3500 (with ATM implied volatility at 28.5%) and up until May
2002 remained in the range 3500-3800, realising around 20% volatility. After May,
the index fell rapidly to around the 2500 level, realising high (around 50%)
volatility on the way. Over the whole year, realised volatility was 36%.
However, the magnitude of the P&L will depend not only on the difference
between implied and realised volatility, but where that volatility is realised, in
relation to the option strike. If the underlying trades near the strike, especially
close to expiry (high gamma) the absolute value (either positive or negative) of
the P&L will be larger.
If volatility is not constant, where and when the volatility is realised is crucial. The
differences between implied and realised volatility will count more when the
underlying is close to the strike, especially close to expiry.
option at an implied volatility below that subsequently realised, and still lose
from the delta-hedging.
VARIANCE
For a clear recap of option’s path dependent volatility exposure: J.P. Morgan, “Variance
Swaps”, 2006, Sections 4.1-4.3.
26
Dollar Gamma: How to make it constant
SWAPS
VARIANCE
27
Remember: A call and a put option with the same
strike have the same gamma (and dollar
gamma). Thus, we can use one or the other.
t 2 St
Our objective is to create a position which provides “pure” (i.e. not path dependent)
exposure to realised variance/vol. We have seen that a single (delta-hedged) option
doesn’t do the work.
In other words, Is there a way of building a portfolio of options such that its
VARIANCE
28
Objective: Constant Dollar Gamma (i.e. constant vol exposure)
Let’s look first at the dollar gamma of different options. We assume a 20% implied vol,
0% interest rates and 1y to expiry.
Dollar Gamma: 50 and 150 strike options Dollar Gamma across strikes
350 50 150 400 25 50 75 100
250 300
250
200
200
150
150
100
100
50 50
0 0
0 50 100 150 200 250 0 50 100 150 200 250
X-axis: stock price. X-axis: stock price.
SWAPS
The dollar gamma of an option has a higher “peak” and a higher “width” as the strike
increases.
VARIANCE
Not quite. Each option, weighted by 1/K, has a similar (peak) dollar gamma, but the
portfolio dollar gamma is not constant with respect to the stock price.
VARIANCE
The area where the dollar gamma of the portfolio is constant depends on the number of
VARIANCE
One important observation is that (peak) dollar gamma increases linearly with strike
(top-left figure next page).
It may be thought that weighting the options in the portfolio (across all strikes) by the
inverse of the strike will achieve a constant dollar gamma. It does have the property
that each option in the portfolio has an equal peak dollar gamma (top-right figure next
page).
However, the dollar-gammas of the higher strike options ‘spread out’ more, and the
effect of summing these 1/K-weighted options across all strikes still leads to a dollar-
gamma exposure which still increases with the underlying (bottom-right figure next
page).
SWAPS
In fact, in can be shown that this increase is linear, and therefore weighting each
option by the inverse of the strike-squared will achieve a portfolio with constant dollar
VARIANCE
32
Dollar Gamma of each option (1) Dollar Gamma of each option (1/K)
Using a subset of options will generate a dollar gamma which is “fairly” constant
on a local area.
34
1/Strike2 – Using options to get a constant dollar gamma
We move from a portfolio of 75 to 225 strike options (25 apart) to a portfolio of 125 to 325
options (50 apart). 1y expiry, 20% vol, 0% rates.
The second portfolio generates a lower dollar gamma, in absolute level, so we will have
to do more notional of each option (or use a finer grid).
VARIANCE
As time approaches expiry, the dollar gamma profile of the portfolio also changes.
As expiry approaches, we will likely need to increase the number of options in our
SWAPS
portfolio to maintain the constant dollar gamma exposure (i.e. use a finer grid of
strikes).
VARIANCE
S 2
t + ∂t − S t
1
P&L =∑ ⋅ Γt ⋅ S t ⋅
2
− σ i2 ⋅ ∂ t
t 2 St
Thus, delta-hedging this portfolio of options will generate a position with a P&L
SWAPS
37
Objective: Constant Dollar Gamma (i.e. constant vol exposure)
If we could buy the entire strike continuum of options, we wouldn’t need to modify
the amount of options.
I.e. static hedge (on the options side; we’ll always have to delta-hedge).
If we assume an initial flat implied volatility skew, the P&L will be just a Flat skew:
implied vol is the
function of realised and implied volatility. same for all
strikes.
S 2
t + ∂t − S t
1
P&L =∑ ⋅ X ⋅ − σ i2 ⋅ ∂ t
t 2 St
When the initial implied volatility is different across strikes, i.e. no flat skew, this will
have an impact given that we buy options with different strikes.
SWAPS
Thus, the “price” of a variance swap will be a function of the volatility skew.
VARIANCE
38
Objective: Constant Dollar Gamma (i.e. constant vol exposure)
Assume we put together a portfolio of (delta-hedged) options with constant dollar
gamma. If ∂t is one business day, i.e. 1/252 years, and we run the trade from day 0
to day T
T
1 S t − S t −1
2
P & L = ∑ ⋅ X ⋅ − σ i ⋅ ∂ t
2
t =1 2 S t −1
X T
St
2
T
1 X T
St
2
T
≈ ⋅ ∑ ln − σ i ⋅ ∑2
= ⋅ ∑ ln − σ i ⋅
2
2 t =1 S t −1 t =1 252 2 t =1 S t −1 252
X T 252 T
St
2
= ⋅ ⋅ ⋅ ∑ ln − σ i 2
2 252 T t =1 S t −1
SWAPS
2 252
39
Objective: Constant Dollar Gamma (i.e. constant vol exposure)
A portfolio of options (calls and/or puts) where each option is weighted by 1/strike-
squared, has constant dollar-gamma;
Hence the P&L from delta-hedging this portfolio is proportional to difference between
realised and implied variance.
This is the idea behind variance swaps: payoff, pricing and hedging.
SWAPS
VARIANCE
Realised
variance
Variance Variance
Seller Buyer
Implied (“agreed”)
variance
SWAPS
Variance swap mechanics ref.: J.P. Morgan, “Variance swaps”, 2006, Section 1.
42
Mechanics
The strike of a variance swap, not to be confused with the strike of an option,
represents the level of volatility bought of sold and is set at trade inception.
The strike is set according to prevailing market conditions so that the swap
initially has zero value.
If the subsequent realised volatility is above the level set by the strike, the
buyer of a variance swap will make a profit; and if realised volatility is below,
the buyer will make a loss. A buyer of a variance swap is therefore long
volatility.
Similarly, a seller of a variance swap is short volatility and profits if the level of
variance sold (the variance swap strike) exceeds that realised.
SWAPS
VARIANCE
43
P&L (I)
The P&L, at expiry, of a (long) variance swap is given by:
Example 1:
An investor wishes to gain exposure to the volatility of an underlying asset (e.g. Euro Stoxx 50) over the next year.
The investor buys a 1-year variance swap, and will be delivered the difference between the realised variance over
the next year and the current level of implied variance, multiplied by the variance notional.
Suppose the trade size is €2,500 variance notional, representing a P&L of €2,500 per point difference between
realised and implied variance.
If the variance swap strike is 20 (implied variance is 400) and the subsequent variance realised over the course of
the year is 15%2 = 0.0225 (quoted as 225), the investor will make a loss because realised variance is below the level
SWAPS
bought.
Overall loss to the long = €437,500 = €2,500 x (400 – 225) . The short will profit by the same amount.
VARIANCE
44
Quotation
Variance swap strikes are quoted in terms of volatility, not variance; but their
payoff is based on the difference between the level of variance implied by the strike
(in fact the strike squared given that the strike is expressed in vol terms) and the
subsequent realised variance.
When quoting and computing the payoff of a variance swap, we won’t use
volatility in % terms; we’ll quote 15% volatility as 15.
Example: you buy a variance swap with a variance notional of 100€ and 15 strike. At
expiry, realised volatility is 20% during the period
Your payoff will be 100€ x (202 - 152 ) = 100€ x (400 - 225) = 17,500€
SWAPS
VARIANCE
45
P&L (II)
Definition of realised variance for variance swap payoff:
2
252 Si
T
σ =2
r ⋅ ∑ ln
T i =1 S i −1
46
Variance Swaps - Recap
Variance swap pays the difference between fixed (implied) and realised variance
Realised variance
variance variance
2
Payout = variance amount x (realised variance - strike ) swap
Fixed Payment
swap
seller buyer
(implied variance = strike2)
Example: Buy €2,500 notional of 6-month variance swap @ 30 strike (variance = 900)
= € 687,500
= € 812,500
VARIANCE
252 2 Si
Realised variance is calculated using the formula :
T i
∑ S
ln
i −1
Taking the first derivative of the P&L of a var. swap w.r.t realised volatility σr we
get 2 ⋅ N Var ⋅ σ r , which depends on the final realised volatility.
Given that final realised volatility is expected to be equal to the swap strike K , a
good approximation to the P&L of the swap for a 1% change in volatility is 2 ⋅ N
Var ⋅ K
2⋅K
r
48
Vega Notional (II)
Using variance or vega notional is irrelevant for the P&L of the swap. However,
market participants will speak in terms of vega notional given that it is related to
volatility, which is the standard measure used in options.
Example 2:
Suppose a 1-year variance swap is struck at 20 with a vega notional of €100,000.
If the index realises 25% volatility over the next year, the long will receive €562,500 = €100,000 x (252 – 202) / (2 x
20). However if the index only realises 15%, the long will pay €437,500 = €100,000 x (152–202) / (2 x 20). Therefore
the average exposure for a realised volatility being 5% away from the strike is €500,000 or 5 times the vega
notional, as expected.
Note that the variance notional is €100,000 / (2 x 20) = €2,500, giving the same calculation as that used in Example
1.
SWAPS
49
N Vega
= N Var
Variance Swaps are Convex on Realised Volatility
2⋅K
Although variance swap payoffs are linear with variance they are convex with realised
volatility.
The vega notional represents only the average P&L for a 1% change in volatility.
A long variance swap position will always profit more from an increase in
volatility than it will lose for a corresponding decrease in volatility (see Recap
example).
This difference between the magnitude of the gain and the loss increases with
the change in volatility. This is the convexity of the variance swap.
If we differentiate the variance swap final P&L w.r.t realised volatility we obtain:
Thus, the sensitivity of the variance swap P&L to volatility is not constant: it is higher
the higher the volatility realised.
VARIANCE
50
Variance is additive
0 t T days
σ 02,t σ t2,T
Realised variance from 0 to t (annualised) Realised variance from t to T (annualised)
2
252 t Si S i
2
∑
T
252
ln
∑ ln
t i=0 S i −1 T −t i =t S i −1
In annualised terms, the realised variance between 0 and T is the weighted
average of the realised variances between 0 and t and 0 and T:
t T −t
σ = ⋅ σ 0 ,t + ⋅ σ t ,T
SWAPS
2 2 2
0 ,T
T T
VARIANCE
51
Mark-to-Market (I)
Marking to market of variance swaps is easy: variance is additive. At an intermediate
point in the lifetime of a variance swap, the expected variance at maturity is simply the
time-weighted sum of the variance realised over the time elapsed, and the implied
variance (i.e. new var swap strike) over the remaining time to maturity.
T −t
MtM t = N Var
t
(
⋅ σ 0 ,t − K 0 ,T +
2 2
) (
K t2,T − K 02,T ⋅ DFt ,T )
T T
SWAPS
annualised realised variance from 0 to t, K 0 ,T was the original (i.e. at time 0) strike, and
K t ,T is the current strike.
52
Mark-to-Market (II)
Example: We are short a 12-month variance swap on a stock
Strike 30%
Variance notional €2,500
Vega notional €150,000 ( = 2 x 2,500 x 30)
Assume that over the next 3 months the stock has a realised volatility of 25% and the
variance swap for the remaining 9 months is quoted at 27.
If we then buy a 9 months var swap with strike 27 and var notional 1,875 ( = 2,500 x 9
/ 12), the P&L would be calculated as :
Variance Amount x ( [Strike2 - realised2] x elapsed time + [Strike2 - Newstrike2] x remaining time )
= €2,500 x [302 - 252] x 3/12 + [302 - 272] x 9/12
SWAPS
= 197 x €2,500
= €492,500
VARIANCE
53
Mark-to-Market (III)
Example 2: Suppose a 1-year variance swap is stuck at 20 with a vega notional of €100,000 (variance notional of €2,500).
If the volatility realised over the first 3 months is 15%, but the volatility realised over the following 9 months is 25%,
then, since variance is additive, the variance realised over the year is:
Variance = [ ¼ x 152 ] + [ ¾ x 252 ] = 525 (≈22.9 volatility). At expiry the P&L would be €2,500 x (22.92 – 202) = €312,500.
Now, suppose again that realised volatility was 15% over the first 3 months. In order to value the variance swap MtM after
3 months we need to know both the (accrued) realised volatility to date (15%) and the fair value of the expected variance
between now and maturity. This is simply the prevailing strike of a 9-month variance swap. If this is currently trading at
25, then the same calculation as above gives a fair value at maturity for the 1-year variance swap of €312,500.
Although the fair value at maturity (now 9 months in the future) is €312,500, we wish to realise this p/l now (after 3-
months). It is therefore necessary to apply an appropriate interest rate discount factor.
If, after 3-months, the discount factor is 0.97, the MtM would be equal to about €303,400.
SWAPS
VARIANCE
These are often set at 2.5 times the strike of the swap capping realised
volatility at this level.
20
20
Final realised volatility Final realised variance
0
VARIANCE
-20 -20
0% 10% 20% 30% 40% 50% 60% 70% 0% 5% 10% 15% 20% 25% 30% 35% 40%
20 strike, 1 vega notional on both swaps
Source: J.P. Morgan. 55
Caps (II)
In practice caps are rarely hit – especially on index underlyings and on longer-
dated variance swaps.
When caps are hit, it is often due to a single large move – e.g. due to an M&A
event or major earning surprise on an individual name, or possibly from a
dramatic sell-off on an index.
Single-day moves needed to cause a variance swap cap to be hit are large and
increase with maturity.
A 1-month variance swap struck at 20 and realising 20% (annualised) on all
days except for one day which has a one-off 14% move, will hit its cap.
A similar 3-month maturity swap would need a 1-day 24% move to hit the
cap
The required 1-day move on a 1-year swap would be 46%.
56
Caps (III)
A 1-y variance swap struck at 20 and realising 20% (annualised) on all days except for
one day which has a one-off 46% move, will hit its cap.
To hit the cap we need:
2 2
T S t +1 252 S t +1
= (2 .5 ⋅ K )
252
∑ ln ∑ ln = (2 .5 ⋅ K )
2 2
where T = 252 and K = 20, i.e.
T t =0 St t =0 S t
We assume that, for all days except one (i.e. 251 days), the stock “realises 20% (annualised)”, i.e.
2 2
252 S S t +1 20 2
1 day “realised (annualised)” variance: ⋅ ln t +1 = 20 2 , which implies: ln = for 251 days
1 S t t S 252
How much does the stock need to change in the other day “m”, i.e. ln(Sm+1/Sm) – as a proxy for (Sm+1-Sm)/Sm , for the
final realised variance to be equal to the cap?
2 2 2 2
T S t +1 S t +1 S m +1 20 2 S
∑ ln = 251 ⋅ ln + 1 ⋅ ln = 251 ⋅ + 1 ⋅ ln m +1 = (2 .5 ⋅ 20 )
2
t =0 S t St Sm 252 Sm
SWAPS
S 20 2
(2.5 ⋅ 20 )
VARIANCE
ln m +1 = − 251 ⋅ = 45 .8
2
Sm 252
57
“Exiting” variance swap positions (before expiry)
Possible ways to exit a variance swap:
If you bought variance with counterparty A, you sell them with counterparty
B; keeping both positions.
Residual risks: counterparty risk if any of the two swaps is not cleared.
SWAPS
58
Offsetting capped variance swaps example
Suppose that an investor buys a 6-month variance swap with a strike of 20. This has
the standard 2.5x cap meaning the exposure to realised volatility will be capped at
50.
The very same day, the (6-month) variance swap trades at a strike of 30 leading to a
significant mark-to-market P&L.
With the strike now at 30, the cap on an new variance swap contract will by default
be set at 2.5 x 30 = 75.
Then if the investors sells this 30-strike variance swap in an attempt to close out his
position the difference in caps will mean he takes on a short volatility exposure if the
subsequent realised volatility is above 50% (although capped at 75%).
In effect, in the course of trying to close out his position, he will have sold a
‘50%/75% call spread’ on volatility. Whilst the price he gets for selling the variance
SWAPS
swap will reflect this higher cap, the residual volatility exposure is presumably
unwanted, and the investor would be best either trading directly with the original
counterparty or negotiating a bespoke contract with another counterparty in order to
VARIANCE
59
Two offsetting var swaps Net payoff position
100 Var. Swap ( strike K = 20% ) P&L Net payoff at expiry
65
Var. Swap ( strike K = 30% ) P&L
80 45
60 25
5
40
-15 Final realised variance
20
-35
0 -55
Final realised variance
-20 -75
0% 5% 10% 15% 20% 25% 30% 35% 40% 0% 5% 10% 15% 20% 25% 30% 35% 40%
10 -35
-10
Final realised variance -55
VARIANCE
-30
-50 -75
0% 10% 20% 30% 40% 50% 60% 70% 80% 0% 10% 20% 30% 40% 50% 60% 70% 80%
Notional on all swaps: 1 vega notional (using the original strike, i.e. 20) – i.e. a variance swap notional of 1 / 2 x 20 = 0.025.
Source: J.P. Morgan. 60
Variance Swap Market
Variance swaps were initially developed on indices.
In Europe, variance swaps on the Euro STOXX 50 index are by far the most
liquid, but DAX and FTSE are also frequently traded.
Variance swaps used to trade on single companies before the Great Financial
Crisis , but they do not trade anymore. Volatility swap trade instead.
The most liquid variance swap maturities are generally from 3 months to around 2
years, although indices and more liquid stocks have variance swaps trading out to 3 or
even 10 years and beyond.
Maturities generally coincide with the quarterly options expiry dates, meaning
that they can be efficiently hedged with exchange-traded options of the same
SWAPS
maturity.
Variance swap market ref.: J.P. Morgan, “Variance swaps”, 2006, Section 2.
VARIANCE
61
Variance Swap Pricing (I)
Pricing a variance swap involves determining its strike price K, i.e. the fixed level
of volatility which will be used to settle the swap (vs. realised volatility) at expiry.
The fair value of the variance swap is determined by the cost, expressed in
volatility terms, of a replicating portfolio.
Our objective here is not to analytically derive how to price variance swaps. Main
references for those interested on that:
1999 Derman et al. (More than you ever wanted to know about volatility
swaps), 2006 Gatheral (The volatility surface), 2006 J.P. Morgan (Variance
Swaps).
SWAPS
VARIANCE
62
Variance Swap Pricing (II)
A variance swap can be replicated by a (dynamically delta-hedged) portfolio of
options with a continuum of strikes weighted by the inverse of the squared strike.
of the entire volatility skew (i.e. implied volatilities for all the option strikes).
Thus, the drivers of variance swap prices are essentially the same drivers as for
VARIANCE
options volatilities and skews (plus particular demand-supply issues on the variance
swap market).
Source: J.P. Morgan. 63
Variance Swap Pricing (III)
A variance swap represents a kind of “weighted average” of volatilities across the
skew curve, with the closer-to-the-money volatilities higher weighted
In the hypothetical case where the skew surface is flat (i.e. all strikes trade at
identical implied volatilities) the variance swap theoretical level will be the
(constant) implied volatility level
This is only true in a simplified world where dividends and interest rates are
zero. In general, the P/L for delta hedging an option is linked to the variability
of the forward, not the spot, while the P/L of a variance swap is a function of
spot volatility.
This leads the variance swap price to differ from what is implied by the static
SWAPS
dividends, interest rates and borrow costs, and their respective correlation to
equities.
64
Variance Swap Pricing - Capped Swaps
A long capped variance swap can be thought of as a standard variance swap plus a
short call on variance, stuck at the cap level.
A standard cap of 2.5x current implied variance strike is relatively far out-of-the-
money.
The pricing of the cap will depend on the volatility of volatility (vol of vol in option
lingo).
Capped vs. Uncapped P&L
A capped variance swap will always be
worth less than an uncapped variance 120 Var. Swap ( strike K = 20% ) P&L
swap of the same strike. Therefore 100 Capped Var. Swap 2.5x P&L
80
capped variance swaps must trade with
60
strikes slightly below their uncapped
40
equivalents – the difference, in theory,
SWAPS
20
representing the current value of the Final realised volatility
0
call on variance.
VARIANCE
-20
0% 10% 20% 30% 40% 50% 60% 70%
35%
30%
As of Dec-10
25%
20%
15%
10%
40 60 80 100 120 140 160
This is due, among other things, to the existence of convexity in the volatility skew,
given that the variance swap price can be thought of as a weighted average of the
entire volatility skew.
SWAPS
VARIANCE
67
Variance Swap Hedging in Practice (I)
A variance swap can be statically hedged with a portfolio of (European-style) options,
weighted according to the inverse squares of their strikes.
This makes it easy, in theory, to perfectly hedge a variance swap with options,
assuming option prices are available across the entire range of strikes.
In practice, traded strikes are not continuous, although for major liquid indices they
are closely spaced (0.4% notional apart for the S&P, 1% for the FTSE, 1.4% for the Euro
Stoxx).
A more serious limitation is the lack of liquidity in OTM strikes, especially for puts,
as these provide a relatively large component of the variance swap price in the
presence of steep put skews.
S&P options are listed down to a strike of 600, FTSE to 3525 and Euro Stoxx
SWAPS
down to 600, although in reality, liquidity does not even reach this far.
VARIANCE
68
Variance Swap Hedging in Practice (II)
In practice, market-makers will not attempt to hedge with the entire strip of options
but typically will use only a few.
One problem with this kind of approach is that the partial hedge is no longer static,
and must be dynamically managed.
Thus, market makers are unable to buy the complete theoretical hedge, and instead
have to use a portfolio comprised of a limited number of options. The resulting
portfolio hedges the variance swap well within a range of asset levels near the spot at
inception, but not outside this range …
SWAPS
See J.P. Morgan, “Variance Swaps“, 2006, Section 4.8for an explanation of how to construct a replicating
VARIANCE
portfolio, i.e. absolute amount of each option traded to generate the variance notional of the variance swap.
69
Variance Swap Hedging in Practice (III)
Assume the client goes long variance and the dealer sets up an (imperfect) replicating
portfolio … In the event that the market falls significantly and realised volatility is higher
than the variance swap strike, the overall hedge will lose money (if it’s not rebalanced).
Dollar Gamma: Var. Swap vs.
replicating options portfolio
Client (long
variance) gets
this P&L
The replicating
hedge gives the Portfolio of options weighted 1/K2 Var. Swap
dealer this P&L (Delta-hedged options with strikes
from 75 to 200, 25 apart.) Stock price
SWAPS
One can imagine what happened in 2008/2009 market crash …For a detailed
explanation, see J.P. Morgan, “Volatility Swaps“, 2010, Section 4.
In order to hedge their variance swap books, market makers were holding portfolios of
single stock options and delta-hedging daily.
We saw in the previous slide what can go wrong if a dealer has sold a variance swap
and hedges it with a partial hedge.
SWAPS
The 2008 crisis led to large drops in single stock prices, and many market makers
VARIANCE
71
2008 Crisis & Variance Swap Hedging (II)
By selling single stock variance swaps, traders had committed to deliver the P/L of a constant
dollar gamma portfolio, irrespective of the spot level, but their replicating portfolio did not
have sufficient dollar gamma at the new spot levels. Market makers were therefore forced to
buy low strike options at the post-crash volatility level, which was much higher than the one
prevailing when they sold the variance swap and therefore incurred heavy losses.
SWAPS
VARIANCE
72
2008 Crisis & Variance Swap Hedging (III)
Not re-hedging the gamma risk was not a possibility, as this would have left the books
exposed to potentially catastrophic losses if the stock prices declined further, and
volatility continued to increase.
This situation led to large losses for many market-makers in the single stock variance
swap markets. In turn this led banks to re-assess the risk of making markets in these
instruments and to a substantial reduction of the liquidity in the single stock variance
swap market.
Index variance swap markets did not experience a similar disruption and were actively
traded throughout the crisis, despite a widening of their bid-ask spreads. Index
variance swaps continued to trade because of the high liquidity and depth of the
index options markets. A wider range of OTM strikes are listed for index options
compared to single stock options. Additionally, the 'gap risk' of a sudden large decline
SWAPS
73
Volatility Swaps
SWAPS
VARIANCE
74
Volatility Swaps
Following the de facto shutdown of the single stock variance swap market in the
aftermath of the 2008 credit crisis, volatility swaps gained liquidity as an instument.
Although pricing and hedging volatility swaps is more complex than variance
swaps, when hedging volatility swaps with options traders are a lot less
exposed to “tail” risks.
There is not a “static” hedge for volatility swaps, thus hedging them requires
dynamically trading options.
P & L = N Vega ⋅ [σ r − K ]
SWAPS
σ
where K is the volatility swap strike, r is the realised volatility and N Vega is
VARIANCE
the vega notional (i.e. P&L for each realised volatility point).
75
Volatility Swaps are Linear on Realised Volatility
A “volatility swap” will have a linear P&L w.r.t. realised volatility, i.e.:
In a vol swap the vega notional is not an approximation to the average P&L if
volatility changes 1%, it is an exact (and constant) amount.
If we differentiate the volatility swap final P&L w.r.t realised volatility we obtain:
76
Volatility Swaps Replication
Whilst volatility can be seen as more of an intuitive measure (being a standard
deviation it is measured in the same units as the underlying), variance is in some
sense more fundamental.
The exposure of delta-hedged options to volatility, after accounting for the dollar
gamma, is actually an exposure to the difference between implied and realised
volatility squared (i.e. variance). In this sense, a variance swap mirrors a kind of
ideal delta-hedged option whose dollar gamma remains constant. Furthermore,
variance swaps are relatively easy to replicate. Once the replicating portfolio of
options has been put in place, only delta-hedging is required; no further buying or
selling of options is necessary.
The main theoretical difficulty with volatility swaps is that the require dynamic
trading in options and therefore a view of forward volatility of volatility.
SWAPS
VARIANCE
77
Volatility Swaps Replication (II)
Delta-hedging options leads to a P&L linked to the variance of returns rather than
volatility. To achieve the linear exposure to volatility (which volatility swaps
provide) it is therefore necessary to dynamically trade in portfolios of options,
which would otherwise provide an exposure to the square of volatility.
There doesn’t exist a “neat” and simple hedging strategy for volatility swaps as it
does for variance swaps (using delta-hedged options with a notional of 1 / strike
squared).
78
No Exam
Stochastic volatility models, such as Heston and GARCH models, have been
proposed by the academic literature.
79
Var vs. Vol Swap Strikes
As opposed to products with linear exposure to volatility, like volatility swaps or
delta-hedged options, variance swaps are convex in volatility and the variance swap
buyer should fairly pay for this convexity, meaning that variance swaps trade above
implied ATM volatility.
Variance swaps trade above ATM volatility because you pay extra for the convexity
of the variance swap: the gain from an increase in volatility is more than the
corresponding loss from a decrease in volatility. This does not come free.
How does the P&L look like with respect to realised volatility at expiry?
VARIANCE
80
Variance vs. Volatility Swaps (I)
Trade: Buy a variance swap & Sell a volatility swap.
Assume first the strike of both variance and volatility swaps are the same
0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100%
If the var. swap strike is equal to the vol. swap strike, you are paying nothing for the
VARIANCE
convexity that the variance swap gives you (vs. the vol. swap), and the trade will be
profitable for sure.
Source: J.P. Morgan. 81
Variance vs. Volatility Swaps (II)
In order to account for the convexity the variance swap provides, variance swap
strikes trade above vol swap strikes. We next show the case of a var. swap strike of
50% vs. a vol swap strike of 40%.
1 vega notional on all swaps
-40 -10
-60 -15
0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100%
SWAPS
The difference between the variance swap strike and the vol swap strike is the price to pay for
the convexity (w.r.t. realised vol) of the variance swap. Thus, trading variance vs. volatility
VARIANCE
swaps is a way of trading the “vol of vol” (the implied vol of realised vol to be precise). Options
on a variance index, like VIX, are another way of trading vol of vol.
Source: J.P. Morgan. 82
Variance vs. Volatility Swaps(III)
Direct link between volatility of volatility and the payoff of the spread of variance
swaps to volatility swaps .
“Spread” defined as a trade where we “buy a variance swap & sell a volatility
swap”.
If volatility of volatility is low then the spread is likely to have a relatively small
payoff at expiry, as the volatility level will likely be close to the strike and the gain
from the convexity of the variance swap compared to the linear volatility swap will be
modest.
In the (unrealistic) limit case where the volatility of volatility is zero, the spread
between the payoffs will be zero.
On the other hand, if volatility of volatility is elevated it is more likely that the
spread will deliver a large payoff at expiry.
SWAPS
Given that the fair value of volatility swaps and variance swaps will reflect these
VARIANCE
dynamics, an increase in the implied volatility of volatility will increase the discount
of volatility swaps relative to variance swaps.
83
Volatility indices – VIX, VSTOXX and related derivatives
SWAPS
VARIANCE
84
Volatility Indices: VIX, VSTOXX, VDAX
The VIX, VSTOXX and VDAX indices represent the theoretical prices of 1-month
variance swaps on the S&P500, Euro STOXX 50 and DAX indices respectively, and
are calculated by the exchanges from listed option prices.
They are widely used as measures of equity market risk, even though they are
only short-dated measures and are not directly tradable.
The short-dated nature of these indices means their level will depend on
perceived futures short-term volatility.
This is important when considering for example event risk that is relatively
distant in the future (e.g. elections).
The design of these indices is based on the square root of implied variance and
incorporates the volatility skew by incorporating OTM puts and calls in the
calculation. A rolling index of 30 days to expiration is derived via linear interpolation
of the two nearest option expiries.
SWAPS
VARIANCE
85
How is the VIX calculated?
The current version of the VIX index was launched in 2003, and measures the 30-day implied variance of S&P 500
options
The first version of the VIX was launched in 1993 and originally measured the 30-day implied volatility of at-
the-money S&P 100 options
The VIX index is constructed by a calculation that is close to that of a 30-day tenor variance swap fair strike
Step 1: Determine which options will be used in the index calculation. We will use only OTM put and call
options of the near-expiry and next-expiry in the calculation
Step 2: Calculate the ‘VIX sub-indices” for the near-term and next-term expiries. T is the tenor of the options
in years, Ki are the option strikes, Pki are the distance between strikes, Q(Ki) is the premium of the out-of-
the-money option of strike Ki and eRT is the discount factor
2 ∆ K i RT1
2
σ Subindex ≅ ∑ 2
⋅ e ⋅ Q (K i )
T i =1 K i
Step 3: The VIX index level is simply the 30-day weighted average of the near-team and next-term VIX sub-
indices
The calculation of the VIX index involves options across the skew, and weights the options contributions
SWAPS
The advantages of using this methodology vs. measuring ATM volatility are:
VARIANCE
86
Example calculation of the VIX index
In the below example, we assume that:
The forwards for the front two expiries are F1 = 920.5 and F2=921.0 and the interest rate is R is 0.38%
The tenors of the two front expiries (in calendar years) are T1= 0.02465 and T2 = 0.10137
The central strike for the two calculations will be K0=920 and we will use options in the range of strikes from
400 to 1220
For simplicity, we ignore the (small) correction factor that corrects for the difference between forwards and
central strikes
The contribution of the 400 put to the near-term sub-index calculation is:
∆ K 400 put
⋅ e RT1 ⋅ Q (400 put ) =
25
2 2
⋅ e 0.0038 ⋅0.0246 ⋅ 0 .125 = 0 .000195
K 400 put 400
The VIX index is simply the weighted average of the two sub-indices – i.e. the square root of the numbers in the
table below
σ Subindex # 1 = 0 . 4727 = 68 . 8 %
σ Subindex # 2 = 0 . 3668 = 60 . 2 %
SWAPS
VIX = 100 ⋅ 61 . 2 % = 61 . 2
VARIANCE
Source: CBOE.
87
VIX and VSTOXX derivatives - futures
VIX and VSTOXX serve as underlying for listed future and option contracts (Eurex
and CBOE)
Futures
Derivatives contract that pay the prevailing (settlement) level of the
volatility index at the expiry date.
Volatility indices are calculated as 1M variance swaps, so volatility
futures allow to trade a future 1M variance swap level.
These futures do not expire on the normal index (futures) expiry dates,
but 30 calendar days beforehand. This expiry is chosen because on that
date, the listed options have exactly 30 calendar days remaining maturity
and the VSTOXX calculation does not need to interpolate from any other
maturities.
SWAPS
References: J.P. Morgan “VDAX, VSTOXX and VSMI Futures”, 2005; Carr &
Wu (2004), “A tale of two indices”.
VARIANCE
88
VIX and VSTOXX derivatives - options
VIX and VSTOXX serve as underlying for listed future and option contracts (Eurex
and CBOE)
Options:
Option contracts (puts and calls) whose payoff at expiry is based on the
settlement (expiry) level of the volatility indices.
These instruments allow trading the volatility of volatility (vol of vol).
In April 2005, options on the VIX index were launched. These represented
the first available exchange traded options on variance. As for the futures,
these expire 30 days before an index expiry and are listed to expire 30
days before the corresponding quarterly options expiry dates for the
underlying.
Reference: J.P. Morgan “Options on implied volatility”, 2010.
SWAPS
VARIANCE
89
VIX facts to know: the spot VIX index is not directly tradable
Fact 1: The spot VIX not directly tradable Fact 2: The beta of VIX futures to spot VIX is not 1
Spot VIX has a constant 30-day maturity VIX futures price in mean-reversion in volatility
The VIX calculation makes the spit VIX index a When VIX is low, the term structure tends to be
‘moving target’, as it referenced a different options upward sloping
portfolio every day When VIX is high, the term structure tends to be
Trading spot VIX would require replicating a constant downward sloping
maturity variance swap
Because of this, longer VIX futures have lower VIX
VIX derivatives provide exposure to forward VIX beta
VIX futures - since Mar-04, settle to spot VIX A longer-dated VIX futures will price in a larger
VIX options – since Feb-04, underlying is VIX future degree of mean reversion, and will therefore be less
sensitive to changes in the spot index (beta lower
VIX ETNs - Provide access to constant maturity VIX
than 1)
futures, leveraged VIX futures, short VIX futures etc.
The spot VIX is a ‘moving target’ while futures are not Beta to spot VIX is higher for shorter maturities
0.6 Beta to spot VIX
0.5
SWAPS
0.4
0.3
VARIANCE
0.2
0.1
Source: J.P. Morgan.
0.0
1M 2M 3M 4M 5M
Source: J.P. Morgan, Bloomberg. VIX futures tenor
90
VIX facts to know: Vol of vol skew and term structure
Fact 3: VIX implied vol is skewed to the call side Fact 4: VIX implied vol term structure is inverted
The VIX spikes when there is a market correction Higher beta for shorter maturity futures
VIX call options can be used for hedging, just like When VIX moves a lot, shorter-dated futures move
S&P500 put options more than longer-dated
Both pay off when the S&P500 fall/implied vol spikes VIX implied vol is related to realised vol of VIX
futures
Demand for hedging/ non-normal return distribution
makes VIX (S&P) call (put) options more expensive VIX implied vol is higher for shorter-dated futures
than put (call) options VIX implied vol term structure is normally downward
sloping
VIX 1M implied vol by strike VIX ATM (wrt. VIX future) implied vol by maturity
80 ATM implied vol
110 1M Implied Vol
100 75
90
70
80
70 65
SWAPS
60 60
50
55
VARIANCE
40
30 50
11 12 13 14 15 16 17 18 19 20 21 1M 2M 3M 4M 5M
Source: J.P. Morgan. Source: J.P. Morgan.
Strike
91
VIX facts to know: implied ‘vol of vol’ risk premium
Fact 5: VIX implied vol is positively correlated with VIX Fact 6: VIX implied vol tends to trade above realised
Large moves in the equity market (mostly to the As with options on the S&P500, VIX implied vol tends
downside) usually lead the VIX to spike; however, to be higher than realised vol of the VIX futures
declines in the VIX tend to be more gradual as there is
no equivalent mechanism that causes the VIX to fall This means there is a vol of vol risk premium
sharply Harvest it by tactically selling options while hedging
This means VIX realised (and hence also implied) vol is VIX spike risk
positively correlated with the level of the VIX
VIX vol is positively correlated to the level of the VIX VIX implied to realised spread
2 0 0%
1 5 0%
SWAPS
1 0 0%
y = 0.0128x + 0.7169
2
5 0% R = 0. 1 6 7
VARIANCE
0%
0 20 40 60 80
Source: J.P. Morgan.
Source: J.P. Morgan.
VIX index level
92
Volatility futures: VIX vs. VSTOXX
VIX vs. VSTOXX futures term structure Spot VIX less VSTOXX
VSTOXX
28 14 2012 avg: 6.7 2015 avg: 7.1
VIX 2011 avg: 5.9
26 12
2013 avg: 4.2
10
24
8 2014 avg: 4.0
22 6
20 4
18 2 Long-term avg: 4.1
0
16
-2
Jan-11
Jan-12
Jan-13
Jan-14
Jan-15
Jan-16
Jul-11
Jul-12
Jul-13
Jul-14
Jul-15
14
Apr 16 May 16 Jun 16 Jul 16 Aug 16 Sep 16
SWAPS
VARIANCE
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
Short Euro STOXX 50 3M var swap - P/L at expiry
-60%
-70%
-80%
Jun-08
Sep-08
Dec-08
Mar-09
Jun-09
Sep-09
Dec-09
Mar-10
Jun-10
Sep-10
Dec-10
Mar-11
Jun-11
Sep-11
Dec-11
Mar-12
Jun-12
Sep-12
Dec-12
Mar-13
Jun-13
Sep-13
Dec-13
Mar-14
Jun-14
Sep-14
Dec-14
Mar-15
Jun-15
Sep-15
Dec-15
Mar-16
SWAPS
VARIANCE
95
No Exam
Investors can use forward variance to trade the future volatility of an underlying.
For a forward volatility position, the P&L before the forward date will be entirely
driven by changes in expectations of volatility, as captured in the implied volatility
term structure.
45%
40%
35%
SWAPS
30%
25%
VARIANCE
20%
15%
Nov-07
Nov-08
Nov-09
Nov-10
Nov-11
Nov-12
Nov-13
Nov-14
Nov-15
Source: J.P. Morgan. 96
No Exam
For example, investors can sell 1Y variance on the S&P 500 and buy 1Y variance on the
Euro STOXX 50.
What will drive the P/L of such a trade? How did this trade perform in the Great
Financial Crisis, in your view?
Long 1Y Euro STOXX 50 var swap and short 1Y S&P 500 var swap
10.0%
Euro STOXX 50 less SPX 1Y var swap levels - vol points
8.0%
6.0%
4.0%
SWAPS
2.0%
0.0%
VARIANCE
-2.0%
Jan-10
Jul-10
Jan-11
Jul-11
Jan-12
Jul-12
Jan-13
Jul-13
Jan-14
Jul-14
Jan-15
Jul-15
Jan-16
Source: J.P. Morgan. 97
References
JPMorgan
“Variance swaps”, 2006.
“Conditional variance swaps”, 2006.
“Volatility swaps, 2010.
“Options on implied volatility”, 2010.
Academic
“Towards a theory of volatility trading”, 1998, P. Carr & D. Madan.
“More than you ever wanted to know about volatility swaps”, 1999, E. Derman et
at.
Carr & Wu (2004), “A tale of two indices”.
“The Volatility Surface”, 2006, J. Gatheral.
SWAPS
VARIANCE
98
Disclaimer
JPMorgan is the marketing name used on research issued by J.P. Morgan Securities Inc. and/or its affiliates worldwide. J.P. Morgan Securities Inc. (JPMSI) is a
member of NYSE, NASD and SIPC. This presentation has been prepared exclusively for the use of attendees at Imperial College “Structured Credit and Equity
Products" Course and is for information purposes only. Additional information available upon request. Information has been obtained from sources believed to be
reliable but JPMorgan Chase & Co. or its affiliates and/or subsidiaries (collectively JPMorgan) does not warrant its completeness or accuracy. Opinions and estimates
constitute our judgment as of the date of this material and are subject to change without notice. Past performance is not indicative of future results. This material
is not intended as an offer or solicitation for the purchase or sale of any financial instrument. Securities, financial instruments or strategies mentioned herein may
not be suitable for all investors. The opinions and recommendations herein do not take into account individual client circumstances, objectives, or needs and are
not intended as recommendations of particular securities, financial instruments or strategies to particular clients. The recipient of this report must make its own
independent decisions regarding any securities or financial instruments mentioned herein. JPMorgan may act as market maker or trade on a principal basis, or have
undertaken or may undertake an own account transaction in the financial instruments or related instruments of any issuer discussed herein and may act as
underwriter, placement agent, advisor or lender to such issuer. JPMorgan and/or its employees may hold a position in any securities or financial instruments
mentioned herein.
99
March 2016
Davide SilvestriniAC
davide.silvestrini@jpmorgan.com
+44(0) 20 7134 4082
This presentation was prepared exclusively for instructional purposes only, it is for your information only. It
is not intended as investment research. Please refer to disclaimers at back of presentation.
CDS Trading Strategies: Credit-Equity Trading
Trading dislocations in credit-equity prices with Equity, CDS and Volatility
Calibration
Sensitivities
Fundamentally, the credit and equity market instruments traded for a company are based on the
same company economic fundamentals
Practically there are sufficient complexities and technicals in each market to provide opportunities
The signals from equity markets to credit markets (and vice versa) are not simple to model or capture
We need a framework of how to link credit and equities
Company Fundamentals
- Earnings
- Cashflow
- Leverage
CDS TRADING
Looking at longer periods for the markets as a whole shows the relationship is more complex than a
simple linear one
STRATEGIES: CREDIT - EQUITY TRADING
Describing the relationship accurately is the challenge, starting with the Merton Model
British Airways Credit Spread and Equity Co-movement Overall Credit Markets versus Equity Markets
Equity price (£, left axis), Credit Spread (bp, right axis Credit = Maggie Credit Industrials Asset Swap Spread
inverted) (bp). Equity = E-Stoxx 50 closing level.
60
5500
Apr 05 - Apr 07
5.5
80 Mar 03 - Apr 05
4500 A B Feb 00 - Mar 03
100
4.5
Jan 99 - Feb 00
3500
120
CDS TRADING
Debt and equity can both be viewed as derivative securities on the underlying company’s asset value
In a Balance Sheet the company’s asset value is equal to the Liabilities (Debt) + Equity
Debt holder is paid (at maturity) as long as the value of the firm’s assets (V) are sufficient to pay the
STRATEGIES: CREDIT - EQUITY TRADING
debt (D), otherwise the debt holders receive whatever asset value there is (recovery rate)
— Debt holder has sold a put on the company’s asset value, with strike D
Residual asset value belongs to the equity holder: Equity holder has bought a call on the company’s
asset value, with strike D
Value
(sold a put)
Equity Equity holder
(bought a call)
Helps explain changing (non-linear) relationship Equity value (S) = max (V-D, 0) Long Call
between credit and equities
STRATEGIES: CREDIT - EQUITY TRADING
400
0 20 40 60 80 100
500
The complexity around this leads us to the JPM CEV
CDS TRADING
In the Merton Model this happens when the Asset Value < Debt Value
So, ‘all’ we need is to calculate for each future point in time are:
The asset value of the company (V)
The volatility of the asset value of the company (σv)
STRATEGIES: CREDIT - EQUITY TRADING
In practice, none of these are observable and so this is not a simple task
Two broad approaches to this have emerged: structural models and reduced-form models
Probability distribution of
x-axis: Time, y-axis: Asset value The Merton
firm's asset v alue
100 Firm asset v alue Framework:
σ (asset value
volatility)
Debt and
80 M ean asset value Equity
(with drift) Holders as
60
Options
40 Positions at
Maturity
20
CDS TRADING
Any model that attempts to accurately Any model that simplifies the modelling
calculate the company’s asset value and process to price credit with fewer inputs
liability structure through time
Do not require detailed inputs of a firm’s
STRATEGIES: CREDIT - EQUITY TRADING
7
Our Requirements For a Useful Debt-Equity Model
It must be theoretically sound
It is not too complex to implement and the inputs should be as simple as is feasible
8
Re-arranging the Balance Sheet Equation to Deal With Market Inputs
We need to imply a probability of defaulting from the equity markets…
But we’ve seen that capturing the point at which unobservable asset value falls below unobservable liabilities is
tricky
We can re-arrange the basic balance sheet equation to try a different approach:
Default will just mean the point at which Equity Price = 0 (when Assets = Liabilities)
Can use observable share price and share price volatility to calculate the probability of this happening
Probability distribution of
x-axis: Time, y-axis: Asset value
firm's asset v alue
100 Firm asset v alue σ (asset value
volatility)
80 M ean asset value
(with drift) Now Using
60 Share Price….
Model of the
40 Firm’s Share
20 Price Through
Time
CDS TRADING
But, the standard Black-Scholes option pricing framework doesn’t allow the equity price to go to zero
Price volatility falls as Share Price falls (volatility is % constant)
So, we use a different option pricing model: the CEV (Constant Elasticity of Variance) Model
STRATEGIES: CREDIT - EQUITY TRADING
This allows % volatility to rise as share price falls, allowing some probability S (share price) goes to zero
ING Group Leverage Effect Equity Volatility Skew for ING Group
x-axis: Share price (€), y-axis: 2m Realised volatility (%) x-axis: Option strike (%); y-axis: 3m implied volatility
35
140
120
30
100
80
25
60
CDS TRADING
40
20
20
0
15
5 15 25 35 45
40 50 60 70 80 90 100 110 120 130 140 150
Source: J.P. Morgan. 11
From Equity Price and Volatilities to CEV-implied CDS Spread
The Roadmap shows the stages to get from equity market inputs The Roadmap
of equity (stock) price and implied volatilities to a CEV-implied 1. Market Inputs
CDS Spread
Equity price Equity implied
A worked example using Dixons (26th Mar 08) will help option v olatilities
Take 70 to 110 strike option prices, for 6m and 1y maturities 2. Solve for σ and α
This creates an equity process consistent
Use the CEV model to calculate a value of σ (= 9.108) and α (= w ith market inputs
0.254)
This gives a probability distribution implied by the stock price
3. Calculate Default Probability
and options markets
Using equity process, calculate
probability equity price = 0, i.e. default
Options Skew and Maturity Inputs
With a standard CDS pricer, we can calculate the CEV-implied CDS Spread from these default
probabilities
100
3Y 9.6% 303.5 185.1 100
If a company is trading at 70bp (5y CDS) in the market and the CEV Implied CDS Spread (using equity
stock and options levels) is 90bp … The market CDS spread is too low (CDS is dear).
Conversely, if the CEV Implied CDS Spread is 50bp … The market CDS spread is too high (CDS is
STRATEGIES: CREDIT - EQUITY TRADING
cheap).
40 50
Cheap CDS: CEV Model
20 CDS - Market CDS = -20bp
0
0 1 2 3 4 5 Source: J.P. Morgan. 14
Comparing our CEV-implied Spread to Market CDS Spreads: Intuition
If the CEV Model Spread > Market CDS Spread…
Position for credit-implied default probability to rise Short CDS risk (Buy protection)
Position for equity-implied default probability to fall Long stock, Short volatility (var swap)
Position for equity-implied default probability to rise Short stock, Long volatility (var swap)
The CDS leg is simpler to isolate default probability… the equity leg needs Stock + Volatility
3% 3%
Probability of Volatility (σ) = 0.9
CDS TRADING
Volatility (σ) = 1
2% Probability of
Default 2%
Default
1% 1%
0% 0%
-15 -8 -3 1 4 5 6 8 11 16 22 -15 -8 -3 1 4 5 6 8 11 16 22 15
Credit-Equity Macro Relationship
iTraxx Main Credit spread vs. Euro STOXX 50 Equity Implied CDS Spreads
5y Spread (bp)
STRATEGIES: CREDIT - EQUITY TRADING
The bursting of the internet bubble in late 2000 shifted corporates’ focus on improving credit metrics,
and led to a period when equity markets priced higher risk premiums than the credit markets.
From mid 2006 to mid 2008, equity and credit markets priced similar risk premiums on an aggregate
level, with the difference between CDS and equity-implied spreads mean-reverting around zero.
CDS TRADING
The Great Financial Crisis led to a very sharp re-pricing of credit vs. Equity risk.
market maker or trade on a principal basis, or have undertaken or may undertake an own account transaction in the financial instruments or related
instruments of any issuer discussed herein and may act as underwriter, placement agent, advisor or lender to such issuer. JPMorgan and/or its employees may
hold a position in any securities or financial instruments mentioned herein.
17
March 2016
Davide SilvestriniAC
davide.silvestrini@jpmorgan.com
+44(0) 20 7134 4082
This presentation was prepared exclusively for instructional purposes only, it is for your information only. It
is not intended as investment research. Please refer to disclaimers at back of presentation.
Equity Correlation Trading
Another tradable dimension
Trading correlation
Correlation swaps
CORRELATION
EQUITY
1
Index Variance as a function of Single Name Variance
Imagine an equity index composed of N stocks with weights wi : i = 1,..., N
Let’s focus on volatility of returns and denote them by σI for the index and σi for
the single names.
You can express the index variance as a function of the single name variances
Example: Imagine a 2 name index with weights w1 and w2 and returns R1 and R2 respectively.
CORRELATION
The index return will be RI= w1 R1 + w2 R2 and its the variance of the index return will be:
Corr (R1 , R 2 )
Var (R I ) = Var (w1 ⋅ R1 + w2 ⋅ R2 ) = (w1 ) ⋅ Var (R1 ) + (w2 ) ⋅ Var (R 2 ) + 2 ⋅ w1 ⋅ w2 ⋅
2 2
2
Introducing Equity Correlations (I)
Index variance is described by the equation:
N
σ I2 = ∑ wi2 ⋅ σ i2 + 2 ⋅ ∑ wi ⋅ w j ⋅ σ i ⋅ σ j ⋅ ρ i , j
i =1 i< j
What happens with index volatility as we reduce stock correlations (i.e. as the
CORRELATION
3
Introducing Equity Correlations (II)
Goal: measure and trade the correlation of the components of an index.
Let’s assume that correlation coefficients are the same for each pair of stocks
ρ i, j = ρ ∀i, j
This may not be realistic, but will facilitate intuition and calculations.
(housing/banks/sovereigns)?
What about during sell-off
triggered by the Tech-bubble
in early 2000s?
EQUITY
4
Introducing Equity Correlations (III)
Assuming a unique correlation parameter:
N
σ = ∑ wi2 ⋅ σ i2 + 2 ⋅ ρ ⋅ ∑ wi ⋅ w j ⋅ σ i ⋅ σ j
2
I
i =1 i< j
N
σ I2 − ∑ wi2 ⋅ σ i2
ρ= i =1
2 ⋅ ∑ wi ⋅ w j ⋅ σ i ⋅ σ j
i< j
CORRELATION
stock volatilities.
5
Introducing Equity Correlations (IV)
Equity correlation as a function of volatilities and weights:
N
σ I2 − ∑ wi2 ⋅ σ i2
ρ= i =1
2 ⋅ ∑ wi ⋅ w j ⋅ σ i ⋅ σ j
i< j
6
Introducing Equity Correlations (VI)
We could either use realised or implied volatilities (both for the index and its
components) and derive two different correlation measures: realised and implied
correlation.
0.80
0.70
0.60
0.50
0.40
0.30
6M ATM Implied Correl 6M Realised Correl
0.20
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
CORRELATION
The existence of a vol surface (skew and a term structure) translates into another
EQUITY
7
Drivers of Market Realised Correlation
A significant driver of correlation between stocks is the prevailing macro
environment.
During periods of high macro uncertainty, stock prices are largely driven by
macro factors such as economic growth, unemployment, interest rate changes,
inflation expectations, etc.
During the inflation and burst of the Technology bubble, stocks were quite volatile,
yet correlation was low due to a strong divergence between stocks in the ‘New
Economy’ (Dot-coms, Technology stocks) and stocks in the ‘Old Economy’ (e.g.,
Utilities, Industrials). This type of intersector performance divergence caused overall
correlation to plummet relative to average stock volatility.
CORRELATION
N
σ I2 − ∑ wi2 ⋅ σ i2
ρ= i =1
2 ⋅ ∑ wi ⋅ w j ⋅ σ i ⋅ σ j
i< j
Assume you think that we are entering a new economic regime where equity
correlations are going to plummet.
CORRELATION
You want to express a short correlation view & limit your net exposure to
volatility.
9
Standard Proxy for Equity Correlation (I)
Equity correlation measures how much stock prices tend to move together. It
provides the link between the volatility of an index and the volatilities of its
component stocks.
Market participants use the following proxy to measure the correlation of an equity
index:
σI ≈ ρ ⋅ ∑ wi ⋅ σ i = ρ ⋅ σ Average
CORRELATION
i
EQUITY
σ Average = ∑ wi ⋅ σ i
i
10
Standard Proxy for Equity Correlation (II)
Using the identity:
2
∑ wi ⋅ σ i = ∑ wi2 ⋅ σ i2 + 2 ⋅ ∑ wi ⋅ w j ⋅ σ i ⋅ σ j
i i i< j
σ I2 − ∑ wi2 ⋅ σ i2
ρ= i
2
CORRELATION
∑ i i ∑ i
w ⋅ σ − w 2
⋅ σ i
2
i i
EQUITY
11
Standard Proxy for Equity Correlation (III)
For a sufficiently large number of components and well-behaved weights and
volatilities, the term ∑ i i
w
i
2
⋅ σ 2
becomes negligible.
This is most straightforwardly observed when the components are equally weighted:
N
1 1 N
σ max
2
∑
i =1 N 2
⋅σ i ≤ 2
2
N
∑ max =
σ 2
i =1 N
N→ 0
→∞
σ I2 σ I2
CORRELATION
ρ≈ =
2
σ 2
∑ wi ⋅ σ i
Average
i
EQUITY
12
Standard Proxy for Equity Correlation (IV)
This approximation makes calculations much easier and straightforward and is quite
accurate for indices with more than 20 stocks and correlation is not very low (>15%
approx.)
SXE5: Index Vol as a function of Correl SXE5: Historical Correl vs. Proxy
Using single stock vols and
weights as of 1-Apr-05
CORRELATION
EQUITY
2
2
σI σI
ρ ≈ =
∑ wi ⋅ σ i σ Average
i
For your reference, some market participants also use what is known as the “Mean
Variance Ratio” (MVR):
σ I2
ρ≈
∑ i i
w
i
⋅ σ 2
CORRELATION
Both proxies result in very similar correlation numbers. The advantage of MVR
has to do with its cleaner relationship with the P&L of correlation trades.
14
Equity Correlations
Index vol should always be less than the average vol of the constituents, due to
the diversifying effects of the index; the size of this discount effectively being
related to the correlation between the constituents of the index.
σI ≈ ρ ⋅ σ Average
Higher correlation means less diversification and hence higher index vol.
EQUITY
σ Average = ∑ wi ⋅ σ i
i
15
Trading Equity Correlation
Short correlation view:
Which instruments should you use
Sell index volatility (options, variance & vol swaps)?
In which notionals?
Buy single name volatilities
The charts below show the daily return of an equally weighted index composed of 5
stocks (A … E) for two different days.
B B
D D
CORRELATION
D D
E E
Index Index
EQUITY
2010
2011
2012
2013
2014
2015
2016
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
CORRELATION
EQUITY
This has historically made short correlation trades profitable. For a backtesting
example see J.P. Morgan “Correlation vehicles”, 2005, p. 28-29.
SX5E 1Y implied less realised Correlation Implied (y-axis) vs. Realised (x-axis) Correl
0.40
90%
0.30 Euro STOXX 50 1Y implied less realised corr
80%
0.20
0.10 70%
CORRELATION
0.00
60%
-0.10
-0.20 50%
-0.40
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
30%
30% 40% 50% 60% 70% 80% 90%
From an analytical point of view: The ratio between index implied and realised
vol has generally been positive and consistently exceeded the ratio between
average implied and realised single stock vol.
ρ imp = σ Iimp / σ Av
imp
. & ρ real = σ Ireal / σ Av
real
.
19
The Correlation Premium (III)
Euro STOXX 50 index to avg SS 6M ATM ratio Implied to realised vol ratio – Index and SS
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
EQUITY
Investors purchasing index puts to protect their downside creates demand for
index volatility, whereas the selling of covered calls (“overwriting”) to earn
alpha increases the supply of single-stock volatility.
Supply of call options via overwriting puts pressure on stock volatility levels.
EQUITY
21
The Correlation Premium (V)
In theory, it is possible for ρ imp = σ Iimp / σ Av
imp
. >1
when correlation is computed
using the standard proxy we have used.
Second, even if our proxy moves above 1, it shouldn’t last very long.
Index implied vol could move above the average single stock correlation in a
scenario where panic causes investors to rush for hedges and the higher
liquidity on the index put option space makes investors use index options
disregarding single stock options.
— These are the same dynamics that can drive the basis-to-theoretical of
an index very very positive (investors hedging with the index and not
with the underlying CDS).
CORRELATION
Even if this happens, this shouldn’t last very long. In a situation like this, we
would expect ...
... Investors establishing short correlation trades
EQUITY
22
Index Correlation & Index Volatility – Relationship
Correlation and volatility (both index and single-stock) are correlated.
Correlation measures how much stocks tend to move together and this is more
likely in times of high vol (e.g. during a sell-off triggered by a negative event).
Euro STOXX 50 1Y correlation and volatility 6M volatility (y-axis) and correl (x-axis)
60%
0.80 1Y realised correlation 1Y realised vol - right axis 45%
Euro STOXX 50
50%
0.70
30% 40%
0.60
CORRELATION
30%
0.50
15%
20%
0.40
10%
0.30 0%
EQUITY
2009
2010
2011
2012
2013
2014
2015
2016
0%
30% 40% 50% 60% 70%
Source: J.P. Morgan. 23
Trading Correlation
There are two principal vehicles for trading correlation:
Correlation swaps
Direct exposure to correlation, paying out on the difference between the
CORRELATION
Vega weighted dispersion trade: the vega notional sold of the index volatility swap
is equal to the aggregate vega notional bought via volatility swaps on the index
constituents.
CORRELATION
25
Trading Correlation: Vega weighted dispersion trade (II)
A vega weighted (long) dispersion trade is short an index vol swap and long single
stock vol swaps, where (Assuming an index vega notional of 1):
The vega amount of each single stock vol swap is equal to the weight of that
stock in the index. P&L of the i-th single stock volatility swap:
P & Li = w ⋅ [σ
i i
real
−σ i
implied
]
The index volatility swap vega notional is the sum of single stock vega notional
amounts, and ∑ w =1
i i
P & LI = 1 ⋅ σ [ real
−σ implied
]
CORRELATION
I I
EQUITY
26
Trading Correlation: Vega weighted dispersion trade (III)
The overall P&L at expiry of the vega weighted dispersion trade is given by:
P & L = ∑ wi ⋅ σ [ i
real
−σ i
imp
]− [σ real
I −σ imp
I ]
i
⇒ σ Ireal = ρ ⋅ σ Av
real
. & σ Iimp = ρ ⋅ σ Av
imp
. ρ real = ρ imp = ρ
In that case, it can be shown that the P&L of the trade is zero only if realised and
implied vol are the same, and that the P&L is positive if realised vol is above implied.
[ ] [ ]
CORRELATION
[
= σ Av
real
− σ Av . −
imp
] [ ρ ⋅ σ Av
real
− ρ ⋅ σ imp
] = [σ real
− σ Av
imp
](
. ⋅ 1− ρ )
EQUITY
. . Av . Av .
27
Trading Correlation: Vega weighted dispersion trade (IV)
Example 1: A vanilla dispersion trade is long volatility
We sell €10,000 vega notional of an index volatility swap at a strike of 20. We buy a total of €10,000 vega notional
of single-stock variance swaps at an average strike of 30.
In order to set up a vanilla dispersion trade the vega notional of each stock will be proportional to its weight in the
index.
index swap.
Therefore (with correlation less than one) a vega weighted dispersion trade (short index volatility, long single-stock
volatility) has a positive vega sensitivity.
EQUITY
28
Trading Correlation: Correlation-weighted dispersion trade (I)
Can we weight the notional of the single stock variance swaps in order for the
variance dispersion trade to have no exposure to volatility (i.e. zero vega)?
Yes, we can.
However, it turns out that, unless we dynamically adjust these weights, we can
only guarantee the trade is vega-neutral at inception
But this already makes the trade much less sensitive to volatility than the
vega weighted one
A volatility exposure will develop as correlation changes
The largest contributor to the P&L of this trade will be the change in
correlation (i.e. implied minus realised) over the life of the variance swaps.
CORRELATION
EQUITY
Which vega notional amounts do we use for the single stock volatility swaps?
29
Trading Correlation: Correlation-weighted dispersion trade (II)
Let’s look at a simple example. Assume:
⇒ σ Ireal = ρ ⋅ σ Av
real
. & σ Iimp = ρ ⋅ σ Av
imp
.
The difference between realised and implied variance for each single stock is
equal to the implied variance times a constant ε (similar for all single stocks):
σ i
real 2
−σ i
imp 2
=σ imp 2
i ⋅ ε , ∀i, ⇒σ real 2
Av . −σ imp 2
Av . =σ imp 2
Av . ⋅ε
Trade:
Short index volatility swap with vega notional 1
Long single stock volatility swaps with vega notional α i ⋅ wi
α = α i , ∀i
CORRELATION
Let’s see if we can find a constant weight scaling factor such that
the P&L of this trade is zero at expiry (given that realised and implied correlation are
the same).
EQUITY
30
Trading Correlation: Correlation-weighted dispersion trade (III)
Trade P&L at expiry of the volatility swaps
[ ] [
P & L = ∑ α ⋅ wi ⋅ σ ireal − σ iimp − σ Ireal − σ Iimp ]
i
[
= α ⋅ ∑ wi ⋅ σ ireal − σ iimp − ] [ ρ ⋅ σ Av
real
. − ρ ⋅ σ imp
Av . ]
i
[
= σ Av
real
. − σ ](
Av . ⋅ α −
imp
ρ )
single name volatility swap) the P&L would have been zero.
EQUITY
31
Trading Correlation: Correlation-weighted dispersion trade (IV)
Trade P&L at expiry of the variance swaps
P&L =∑
i
α ⋅ wi
2 ⋅ σ iimp
⋅ σ [
i
real
− σ i
2
imp
−
1 2
2 ⋅ σ Iimp
]⋅ σ real
I − σ[imp
I
2 2
]
= α ⋅∑ w ⋅
[
σ real 2
Av . −σ imp 2
Av . ] − [ρ ⋅ σ − ρ ⋅σ
real 2
Av .
imp 2
Av . ]
2 ⋅ σ Av 2 ⋅ ρ ⋅ σ Av
i imp imp
i . .
σ Av . ⋅ε
( )
imp
= ⋅α− ρ
2
single name variance swap) the P&L would have been zero, independently of ε (i.e.
Whether realised vol ends up being lower or higher than implied vol).
EQUITY
32
Trading Correlation: Correlation-weighted dispersion trade (V)
The trade is initially vega neutral. As implied correlation moves, in order to keep the
trade vega neutral the investor would theoretically have to rebalance the vega
notional of the single name volatility swaps (in theory, at least).
As correlation increases the long dispersion (i.e. short correlation) trade becomes
short vega.
Why?
Initially we are short index vol with a vega notional of 1 and long single
name vol with a total vega notional of ρ imp
If implied correlation increases, then under the new implied correlation our
initial total notional on the single name swaps is too low.
CORRELATION
Given that correlation and volatility are positively correlated, this typically works
against the trade P&L. This explains the negative convexity of the trade.
EQUITY
33
Trading Correlation: Correlation-weighted dispersion trade (VI)
Example 2: A correlation-weighted dispersion trade is initially vega neutral
Consider index volatility at 20%, average singe-stock volatility at 30% and correlation at approximately 0.44.
For a vega weighted dispersion trade the weighting of any single-stock volatility swap in the trade (in vega-notional
terms) would be the weight of the stock in the index.
To compute the weighting for the correlation-weighted dispersion trade we multiply this stock weight by the square
root of implied correlation (i.e. 0.66).
Therefore (assuming all stocks have volatilities equal to the average single-stock volatility) we will be short 3x index
volatility and long 2x single-stock volatility (i.e. if index vega notional = 1 and total single stock vega notional =
0.66).
Suppose as before that single-stock volatility increases by 1% whilst correlation stays constant.
Then according to the Example 1, index volatility will increase by 0.66% whilst single-stock volatility will increase by
1%. Since we are short index to single-stock volatility in a 3:2 ratio our net P&L will be zero.
Suppose instead that correlation decreases by 4 correlation points whilst stock volatilities stay constant. Our trade
will still be weighted in a 3:2 ratio. Single-stock volatility will remain unchanged, but, using the usual correlation
CORRELATION
proxy formula, index volatility will increase to 30% multiplied by the square root of 0.40, equal to 19%. Therefore,
since we are short index volatility, our trade will make a profit of 1 vega, as a result of the favourable move in
correlation (given our index notional vega of 1 and the fact that single stock vol didn’t change).
EQUITY
34
No Exam
The swap on each single name has a vega notional equal to the weight of that
name on the index (times the index vega notional). As a consequence, the sum
of the vega notionals of the single name swaps adds up to the vega notional of
the index swap.
A long vega weighted dispersion trade is short correlation and long volatility.
The sum of the vega notionals of the single name vol swaps adds up to the vega
notional of the index swap times the square root of correlation.
36
Trading Correlation: “Enhanced” Dispersion Trades
Trading variance swaps in all the index constituents may not be the most efficient
way of implementing a dispersion trade given:
37
Exercise
Trade: long dispersion (short correlation) using volatility swaps
Assumptions:
1. Equally weighted index made up of 2 single stocks (i.e. 50% weight each).
2. Vol swap strikes equal to implied vol for all index and single names.
3. Implied vols for the two single names are always the same.
4. Initially single name implied vols are 30 and the index implied vol is 20.
5. Interest rates are zero, i.e. discount factor is 1.
You are asked to compute the MtM of this trade in different scenarios for changes in implied vols and implied
correlation, assuming they change the very same day the trade is opened.
Remember that the MtM of a (long) volatility swap is equal to: Vega Notional x (New – old strike) x DF. In our case,
DF = 1 and strikes are equal to impled vols.
A. Vanilla dispersion trade: single name stock volatilities increase from 30 to 31 and implied correlation remains
constant.
B. Correlation-weighted dispersion trade:
CORRELATION
B1. Single name stock volatilities increase from 30 to 31 and implied correlation remains constant.
B2. Single name stock volatilities stay at 30 and implied correlation decreases to 0.4.
B3. With the new implied correlation (0.4) and assuming you do not rebalance the vega notionals of your single
stock volatility swaps, assume single name stock volatilities increase from 30 to 31 and implied correlation
remains constant (at 0.4).
EQUITY
38
Trading Correlation: Correlation Swaps (I)
Correlation swaps give direct exposure to correlation, paying out on the
difference between the strike of the swap and the subsequent realised average of
pair-wise correlation of a pre-agreed basket of stocks.
Realised
correlation
Correlation Correlation
Seller Buyer
Implied (“agreed”)
correlation
CORRELATION
Notice that the measure of correlation here is pair-wise correlation of (the daily
EQUITY
39
Trading Correlation: Correlation Swaps (II)
The P&L, at expiry, for a (long) correlation swap is given by:
40
Trading Correlation: Correlation Swaps (III)
Advantages of correlation swaps:
The correlation swap gives direct exposure to the level of delivered (average
pairwise) correlation with no dynamic hedging/replication required (for the
investor; obviously the dealer will have to hedge the exposure!).
41
Sources for “correlation” data (I)
The realised correlation of a pair of stocks measures how much the stock prices tend
to move together. The realised correlation of an index is simply an average across all
possible pairs of constituent stocks. There are two ways of computing this average:
The advantage of using the index realised correlation measure is that it is possible to
use the same method to back out an implied correlation measure from the implied
CORRELATION
index and single stock volatilities. This implied correlation measure represents the
market’s expectations of future realised index correlation inherent in the implied
volatility. Note that we can only compute this implied correlation if the index upon
which it is based has actively traded volatility/options.
EQUITY
42
Sources for “correlation” data (II)
Since the payoff of a correlation swap is equal to the average pairwise correlation and
this is usually very close to the realised index correlation, the strike of a correlation
swap should in theory trade close to the implied correlation described above.
43
Risks of Trading Correlation
Correlation moves with volatility:
Factors which cause volatility to increase tend also to lead to an increase in
correlation. So whilst a correlation swap or correlation-weighted dispersion
trade (at least initially) is technically vega-neutral, if volatility were to
increase, it is likely that correlation would also increase.
In this case, a long correlation-weighted dispersion trade, which is short
correlation, would become short vega also and the losses would be magnified.
In the same way a decrease in correlation would most likely be accompanied by
a decrease in volatility which would act to limit the potential gains. This
generally makes the P&L of correlation trades negatively convex on realised
correlation.
Reconstitution risk: When a dispersion trade is set up the volatilities of the current
index constituents are traded against the volatility of the index. Over time it is
CORRELATION
possible that the index members (and/or their weightings in the index) will change.
Index members may be demoted, may be bought out by other companies, or may
move their listing elsewhere. In this situation the basket of single-stock variance
swaps held may cease to be representative of the index and, without re-weighting,
the hedge against volatility will partially break down. Note that this risk will increase
EQUITY
44
References
JPMorgan
45
Disclaimer
JPMorgan is the marketing name used on research issued by J.P. Morgan Securities Inc. and/or its affiliates worldwide. J.P. Morgan Securities Inc. (JPMSI) is a
member of NYSE, NASD and SIPC. This presentation has been prepared exclusively for the use of attendees at Imperial College “Structured Credit and Equity
Products" Course and is for information purposes only. Additional information available upon request. Information has been obtained from sources believed to be
reliable but JPMorgan Chase & Co. or its affiliates and/or subsidiaries (collectively JPMorgan) does not warrant its completeness or accuracy. Opinions and estimates
constitute our judgment as of the date of this material and are subject to change without notice. Past performance is not indicative of future results. This material
is not intended as an offer or solicitation for the purchase or sale of any financial instrument. Securities, financial instruments or strategies mentioned herein may
not be suitable for all investors. The opinions and recommendations herein do not take into account individual client circumstances, objectives, or needs and are
not intended as recommendations of particular securities, financial instruments or strategies to particular clients. The recipient of this report must make its own
independent decisions regarding any securities or financial instruments mentioned herein. JPMorgan may act as market maker or trade on a principal basis, or have
undertaken or may undertake an own account transaction in the financial instruments or related instruments of any issuer discussed herein and may act as
underwriter, placement agent, advisor or lender to such issuer. JPMorgan and/or its employees may hold a position in any securities or financial instruments
mentioned herein.
46