FIXED INCOME AND CREDIT RISK

Course description:

 Classical theories of the term structure of interest rates

 Stochastic interest rate modeling (arbitrage free term structure models)
 Interest rate risk measurement and management : duration and convexity (bullet bonds as well as bonds with
embedded options)
 Immunization and hedging
 KRD (key rate durations) and non-parallel shifts of the term structure
 Analysis and valuation of floating rate notes and interest rate swaps
 Analysis and management of credit risk : credit risk models, DTS (duration times spread) and exposure to spread risk
 Use and valuation of credit default swaps (CDS)

General plan:

1. Classical theories of the term structure of interest rates

2. Stochastic models of interest rates
3. Systematic risk, duration, convexity & key rate durations
4. Floating rate notes
5. Interest rate swaps
6. Credit risk models
7. Duration times spread & the credit risk exposure
8. Credit default swaps

Lecture 1
The term structure of interest rate experiences mostly non-parallel shift. Floating rate notes were based on 3-month interest
rates like the LIBOR and they transitioned away from the LIBOR: how does it influence/affect the valuation of floating notes
today as the reference interest rate today is an overnight interest rate. Issuers of floating rate notes that are not risk free
compared to the reference interest rate will expose you mostly to credit risk with a very interesting consequence: we can be
exposed to negative duration.

IRS + duration time spread + CDS: tool that makes it possible to manage interest rate risk.

Chapter 1: Classical theories of the term structure of interest rates

1. Spot rates and the term structure of interest rates

 Spot rates
 Term structure of interest rates
2. Spot rates and forward rates
 Forward rates
 Implied forward rates
 Expected future spot rates
3. Classical theories of the term structure
 Expectation theory: why do we have a given shape of the term structure of interest rates and how it evolves
over time. What makes an equilibrium for bond markets in different maturities? What makes the equilibrium
in terms of interest rate markets for different maturities.
 Liquidity theory: we consider the fact that there is a risk premium that the market requires
 Market segmentation theory: Strict segmentation + Preferred habitat theory

We observe the price of bonds that are zero coupon bonds, not the spot rates. The price of the bond is the sum of the
discounted cashflows. A zero-coupon issued by the US/French/German Treasury is called STRIPS (Separate Trading of
Registered Interest & Principal Securities). In US the interests are paid twice a year (semiannual coupons). The STRIPS gives
the chance to trade the cash flows separately. A coupon bond is a portfolio of zero-coupon bonds. The 5-year US Treasury
Coupon Bond is a portfolio made up of 11 different zero-coupon bonds: 10 coupons and one that is the redemption price. We
can trade separately the different cash flows of a coupon treasury bond. The first cash flow is a six months zero coupon bond
and the tenth cashflow is a 5-year zero coupon bond. We have a zero-coupon bond for each quarterly maturity up to 30 years:
we have a very nice idea of what the spot rates are, from 6 months up to 30 years and we have an information on the spot rates
for each quarterly maturity up to 30 years.

Term structure of interest rates or term structure: the set of spot rates for different maturities in t=0, for a homogeneous
class of securities.

With STRIPS, for each maturity we have a spot rate, and the set of spot rates is the term structure of interest rates for the US
Treasury market.

Main shape of the term structure:

 Upward sloping (positively sloped): the further the maturities the higher the spot rate. In most of the cases
  Downward sloping (negatively sloped or inverted): generally, after a long period of expansion we witness inflationary
tensions or pressures, the economy starts to overheat and suddenly the CB senses that is time to rise the short rates
to cool the economy.
 Flat term structure of interest rates: it is very rare. It means that the level of spot rates is the same for all the different
maturities. This is quite rare and is rather a theoretical case, which is used in academics.
 Humped: for shorter maturities we have rising spot rates and then with longer maturities spot rates are lower and lower
 Humped back: spot rates that first decrease with maturity and then they increase with longer maturities.

On the US market we have the STRIPS, so we can calculate the spot rates for each quarterly maturity up to 30 years.
When we don’t have zero coupon bond is a problem.

When zero-coupon bonds are not available: bootstrapping to extract spot rates from coupon bonds or IRS rates (swap
transactions)

o Interpolation methods:
o Linear interpolation
o Cubic interpolation
o Cubic differential interpolation

Each cash flows should be discounted based on the spot rate. If we have a one-year coupon bond and a 2-year coupon bond
we can extract the 2-year spot rate. More and more today different statistical methodologies are used to estimate the term
structure and this estimation is based on all the available information of all coupon bonds and short-term bills available for a
given class of risk or for a given bond market.

There are mainly two approaches in term of term structure estimation:

 Exponential spline fitting (Vasicek & Fong)

 Parametric Modelling: used more and more by CBs

It is not so easy to model and obtain the term structure of interest rates. Why don’t we just use the yield curve as a substitute for
the term structure of interest rates? Most of the bonds are coupon bonds why we don’t just calculate the yield to maturity of the
different coupon bonds for the different maturities and establish a yield curve? This is not a good idea. Yield to maturity is a
biased estimator of the spot rate. Yield curve cannot be a substitute for the term structure of interest rates. The yield curve
cannot be a substitute of the term structure of interest rates. The is only one case where the yield curve is an unbiased
estimator of the term structure of interest rate: when the term structure is flat. if we consider the 10-year maturity in an upward
slope structure of interest rates the yield to maturity will be lower than the spot rates. In downward slopes of interest rates, the
yield to maturity will be higher than the spot rate. The price of a coupon bond maturing in n years is the sum of the discounted
cash flows, each cash flow being discounted based on the spot rate for the given maturity. The yield to maturity is a complex
weighted average of the spot rate. Yield to maturity is a biased estimator of the spot rate and the bias will be the more important
the larger the coupon. The lower the coupon the smaller the bias. When the coupon tends towards zero, the YTM will be equal
to the spot rate

Lecture 2
Spot rates & Forward Rates

Investment strategies over n years horizon:

 Maturity strategy: invest over n years considering n year spot transaction.

 Rollover strategy: strategy where you borrow a first time for one year, invest for one year and then you will
successively reinvest the product of the investment at the successive forward rate. After one year I will reinvest (1+s 0,1)
at the forward rate h1,2.

Implied forward rates

What the level of the quoted forward rate should be. We can always use spot rates to determine what level of the forward rate
should be. Forward lending transaction. If the forward market doesn’t exist, we can synthetically build the forward transaction.

 S0,3/12 à spot rate for 3 months zero coupon.

 F2,4: implied forward rate (IFR) for a forward transaction to be carried out in two years and maturing in 4 years.
 F3/12,6/12: forward rate carried out in 3 months and maturing in 6 months.

The IFR gives us the level (in the risk-free environment) towards which the forward rate should tend. We have the term structure
of the 1-year forward rate. From the same spot curve we can calculate different term structure of forward rate. Is there a link
between the forward rates and market expectations regarding the future level of spot rates? In a frictionless world h1,2 should
tend to f1,2 by arbitrage. We can build a synthetic forward transaction. We lend, in the case below, at h1,2 and we borrow at
f1,2. We finance a lending transaction at h1,2 rate with a forward borrowing transaction at f1,2 rate.

There is a link between the observed spot rate and the implicitly observed implied forward rate and the expected future spot
rate. This is a rollover strategy with implied froward rates. The expected future spot rate can be considered a set of short rates.
The long rate can be considered as the product of a set of short rates (the one-year rate and all the implied one year forward
rate).

Is there a link between the implied forward short rate and the expected future short spot rates? If we accept that there is a link
between the implied forward rate and the expected future spot rate then there will be a role which is played by the expectations,
in the shape, in the level, in the evolution of the term structure and in the market equilibrium. This is the focus of the classical
theories of the term structure à provide one explanation of the shape, the level and the evolution of the term structure (the
market equilibrium). Is there a link between fn-1,n (the implied forward rate) and the expectation of the future short term sport rate?
Each longer-term spot rates can be seen as the product of a set of short-term interest rates. Agents’ expectations could have a
role in the shape and level and evolution of the term structure of interest rates. Expectations will play a major role in market
equilibrium. This is the core of the classical theory of term structure of interest rates.

These theories consider the nature of the expectations, the agent preferences and the attitude toward risks and they try to
provide a rational explanation to market equilibrium. Most of these theories are aimed at explaining the link between the spot
rates of different maturities and the expectation regarding the future level of these rates.

There are several theories of expectations. They tell us that the term structure of interest rates reflects the average, the
consensual view of the market, of the agents, regarding the future level of interest rates. This means that if we have an upward
sloping term structure in the strictest version the expectation theory states that an upward sloping term structure implies that
agents are expecting an increase in future spot rates and the downward sloping term structure of interest rates implies that
agents say that interest rates will go down, In its strictest form we have the pure expectations theory. It states that the implied
forward rate is equal to the expected future spot rate. Pure expectation theory: whatever the investment strategy it will yield the
same expected total return over given investment horizon. The maturity strategy, the rollover strategy or a buy and hold will
provide the same expected total return. They are risk neutral, they are indifferent to reinvestment risk, indifferent to price risk. In
the pure expectation theory, whatever the investment strategy, this investment strategy will yield the same expected total return
on a given investment horizon. In market equilibrium all the bonds are priced in such a way that on a given investment horizon,
whatever the investment strategy we will get the same expected total return. An upward sloping term structure implies that the
market expects an increase of the future spot rates and a downward sloping term structure implies that the market expects
decreasing future spot rates. How does the term structure change? Having an expectation lager than the implied forward rate: is
it compatible with equilibrium? If there is no equilibrium as investors, we will favor the rollover strategy and as borrowers we will
favor the maturity strategy. How can we go back to equilibrium if the expectation of the market really is higher than the IFR?
Rollover strategy for investors and maturity strategy for the borrowers. The rollover strategy is more attractive for the lenders
investors and less attractive for the borrower
We can suppose that the real interest rate is stable over time, the nominal rate or the inflation expectations will move (higher
inflation rate à nominal rate should increase). The expectation theory explains all kind of shapes (flat, upward sloping,) but the
limits are numerous. The first problem is that the theory doesn’t explain why agents are expecting an increase/decrease of
future spot rates. We should link the expectation of the agents to the expectations of future inflation: inflation expectations. The
real interest rate is stable over time. If we look at the historical terms structure of interest rate: it is mostly upward sloping. On
average we expect systematically an increase in inflation rate: on average we are expecting higher inflation. Agents are
supposed to be risk neutral: they are indifferent to price risk and reinvestment risk. The rollover strategy implies reinvestment
risk. In the pure expectation theory agents are risk neutral: they are not concern by price risk and reinvestment risk: the only
thing that matters is to maximize their expected return. In the pure expectation theory there is a very disturbing assumption:
statistical independence (there is no correlation between the future spot rate). The correct formula is the one that considers the
expectations. This formulation is only correct if we suppose that there is no correlation between the future spot rates, which is
rather restrictive assumption. There is incompatibility between the different versions of the expectation theory. The liquidity
theory introduces liquidity premium to encourage lenders to consider longer dated maturities.

Lecture 3
CDS are used to manage credit risks. We considered two different strategies:

 Maturity: over a horizon of n years. We are investing at the spot rate S0,n à we are buying the zero-coupon bond
maturing in n years.
 Rollover: we invest in the short spot rate S0,1 for one year and renewal year after year at the quoted one year forward
rate

We have market equilibrium if the returns are equal. If the market is sufficiently liquid, we have good chances to exploit the
mispricing among the two different strategies. The forward rate market is not necessarily liquid for all the maturities so it's not
always possible to have this equality. When we cannot observe the quoted observed forward rate, we have an interesting
indication if we can observe the spot rate (i.e. s0,1 s0,2). We can calculate the implied forward rate.

We can create synthetically the forward transaction and the rate linked to this transaction is the implied forward rate.

(INSERT EXERCISE 1)

We went through the different assumptions and limits of the pure expectation theory. It is simple and quite attractive: it explains
all kind of shapes of the term structure but there are limits. It doesn’t explain agents expect an increase/decrease of future spot
rates à we introduced the expectation of inflation. Real rate component is relatively stable over time and independent from the
level of inflation. In the pure expectation theory, agents are risk-neutral and they are indifferent to price risk as well as to re-
investment risk. There is incompatibility of the different versions of the expectations theory. The local expectation theory is the
only one considered compatible with the market equilibrium and it makes a very simple assumption: whatever the maturity of
the bonds that we are considering, all the bonds are offering the same expected return over the short-term: they limit the
analysis over 1 period and it avoids the comparison between different investment strategies.

Using the current spot term structure and they extract the implied forward rates and it is the basis for building the market
expectations in all the different asset classes.

 f: implied forward rate

 h: observed forward rate

The phases of economic expansion are much longer than the phases of economic contraction. In the unbiased version of the
expectation theory (i.e., the pure expectation theory) there is an assumption that there is no correlation between the future spot
rates. This is a quite restrictive assumption. We have 4 versions of the theory and they are not really compatible among them.

Local expectations theory: it is the only one compatible with market equilibrium. The local expectations theory reduces the
problem of the compatibility, considering that the pure expectation theory works well on the short term. Pure expectations
theory: the bond maturity plays no role in the investment decision. Can the pure expectation theory really be used to forecast
future spot rates? If we believe in the pure expectation theory the shape of the term structure, that we can observe today, will tell
you what the rates should be for all the different maturities. We want to understand the difference between the Actual Future
Spot Rates and the Implied Forward Rates

Liquidity theory
In the liquidity theory agents are risk averse. Lenders are averse to price risks and when they consider a longer-term
investment, they will prefer the rollover strategy rather than the maturity strategy. If they have to liquidate their investment before
the end of their time horizon, they expose themselves to price risk. The lenders will require a risk premium to invest longer
terms. In the liquidity theory borrowers favor the long term, they are averse to reinvestment risk. The borrowers are up to paying
a risk premium to borrow over the long term. Lenders are risk averse and borrowers are reinvestment risk averse. The maturity
strategy must generate a higher expected strategy than the rollover strategy. Under the liquidity theory the IFR is higher than the
expected future spot rate

Lecture 4
Pure expectation theory: mechanism on how new information is liable to change the shape of the term structure by an arbitrage
process. In the PET we assumed that agents are risk neutral: indifferent to price risk as well as re-investment risk, they just want
to maximize the expected return. In the liquidity theory the investors prefer the short term. When they are considering longer
dated investment horizons, they will favor rollover strategies vs the maturity strategy. With the maturity strategy we are exposing
ourselves to price risk. There are two elements of the interest rate risk/systematic risk that affect all bonds investments: price
risk and reinvestment risk. When interest rates go down the reinvestment risk means that each coupon will be reinvested at a
lower interest rate. Each coupon then will be reinvested at a lower rate. If we sell a bond when the interest rate goes up, we will
have a capital gain. Investors are risk averse towards the price risk. We require a risk premium when buying longer dated
bonds: the borrowers will pay for this premium.

In the liquidity premium theory the implied forward rates aren’t anymore an unbiased estimator of the expected future spot rates.
The liquidity premium is supposed to grow whit the maturity as the theory simply states that the premium diminishes and
vanishes as time flows by. The liquidity premium is growing with maturity but at a digressive rate.

If you know the IFR and the liquidity premium it appears that the expected future spot rate is equal to the IFR – the liquidity
premium. As the IFR is not anymore an unbiased estimator of the expected future spot rate we can have a positively slope term
structure of interest rates and an expectation of declining future spot rates. Under the liquidity theory we have 3 cases
considering 2 periods, given the shape of term structure and the direction of expectation.

An upward sloping term structure is compatible with the expectation of decreasing future spot rates. The liquidity theory
completes the pure expectation theory by introducing the agents’ attitude toward risk. The liquidity premium theory gives some
explanation to the fact that we could have more frequently upward sloping term structure even though people are expecting a
decrease in future spot rates. The liquidity premium is positive and increasing with maturity
The last theory is the market segmentation theory:

 Strict segmentation theory: there is no unique market on which all maturities are traded

According to this theory (strict segmentation theory) there are as many markets as there are investment horizons and maturities.
The strict segmentation introduces the absolute risk aversion

 Preferred habitat theory of the segmentation theory

It accepts the idea that the term structure contains information on expectations as to the future spot rates and on risk premia:
any agent can consider adjusted maturities as long as the risk premium is sufficient. This theory refuses the absolute risk
aversion of the strict segmentation theory.

The PHT says that the IFR is equal to the expected future spot rate + a liquidity premium can be positive/zero/negative. This
can explain every shape of the term structure, it is a very general theory of market equilibrium but at the same time is very
limited.

The predictive power of this theory is very limited and the theory doesn’t say much about the level and on the behavior of the
risk premium and it has limited predictive power. If the premium tends toward zero, we find ourselves with agents becoming
indifferent to these types of risks (reinvestment risk and price risk). When the premium goes to zero the IFR becomes an
unbiased estimator of the future spot rate. And we are in the field of the pure expectation theory. When the premium is
systematically positive for all the different maturities, we find ourselves again in the liquidity theory.

Chapter 2 now

Stochastic model of interest rates

Excel: binomial tree of interest rates to value almost any type of interest rate instruments.

 One-factor stochastic models

 Arbitrage-free models
 The binomial interest rate tree

Binomial tree with mean reversion: log-normal model is one of the simplest models. We add the mean reversion. Stochastic
models make it possible to price interest rate risk.

Most of the models are built on the short-term interest rates and they consider the statistical characteristics of short-term interest
rate movements:

 Drift: expected direction of the change in the short rate

 Volatility
 Mean reversion: tendency to come back to the average long term mean of the short rate
We can integrate these characteristics in SDE (stochastic differential equations). The short-term interest rate follows a
stochastic process. The shift factor (parallel movement in the term structure of interest rates) explains most of the movements of
the term structure. Two factor models are more difficult to implement.

Continuous-Time Stochastic Process

(2.1) is a generalized Wiener process. When b=0 e sigma=1 the continuous-time stochastic process is a standard wiener
process. The stochastic nature of dr is directly linked to the random process dz. The problem with this equation is that the drift
and the volatility aren’t function of the underlying variable r at time t à we have to rewrite the equation in order to specify the
drift and the volatility term.

Equation (2.2) means that the 2 terms are function of the value of the short-term rate r at time t. This equation is called Ito
process and it reminds a Markov process because it considers the change in r at time t.

The change depends only on the value of the short rate at time t. Only the current value of r is relevant to predict the future. A
Markov process is a particular type of stochastic process in which only the current value of a variable is relevant to predict the
future. The distribution of the future at any given future time doesn’t depend on the particular path that was followed by the
variable in the past. Only the current value is relevant. The specification of the stochastic process makes us introduce a
statistical characteristic of interest rate: we can integrate the mean reversion in the drift term. When we introduce mean
reversion, we won’t have the same binomial tree over the same time horizon: we won’t have the same binomial tree than we
accept the volatility is constant for example.

For valuing fixed income instrument we will use arbitrage-free models. The models are adapted to the current term structure of
interest rate. The models will provide prices for the bonds that corresponds to the observed market prices on the market.

o Ho-Lee: short rates are normally distributed

o Hull-White: short rates are normally distributed and with mean reversion
o Kalotay-Williams-Fabozzi: short rates log-normally distributed (we cannot have interest rates)
o Black-Derman-Toy: short rates log-normally distributed and endogenous mean reversion determined by non-constant
volatility (many people use this one)
o Black-Karasinski: short rates log-normally distributed and mean reversion
o Heath-Jarrow-Morton: multifactor model

We will focus on the log-normal model. Another version of the model is called the shifted log-normal model. We introduce the
shift, which makes it possible to use the model notably to price interest rate derivatives.

When in the BDT model we consider constant volatility, we end up with the KWF model.
We have the binomial tree or the trinomial tree in the HW model.

We will build the binomial lattice with the lognormal model that has been developed by Kalotay-Williams-Fabozzi. At each node
of the lattice the model assumes that the short-term rate evolves over time according to a lognormal random walk with a
constant volatility, we assume that the volatility is constant. At the base of the lognormal model we assume that the volatility is
constant. The natural logarithm of the short rate is normally distributed: the short rate is lognormally distributed

We have constant volatility and we have a constant drift.

KWF model = lognormal model, we have to interest ourselves in the relation between the higher and the lower rate.

We would like to know all the different possible level of the short rate over 4 years. Priced at par: the YTM is equal to the
coupon bond.

The YTM of a bond priced at par is defined as the par yield.

One element of the tree: R0, is the one-year spot rate. At each node I can value the price of the bond, knowing what the rates
are. At each node I can calculate the value of the bond because there are 2 possible rates (one up and one down).

We want to find the lower and higher rate in year 1. To do that we will use the 2-year bonds. At the end we have the redemption
price and the coupon. We know the coupon, the redemption price, and the volatility: there is only one variable. We have the
relationship between the higher rate and the lower rate.

The simple analytical solution is to solve the quadratic equation: we use a numerical solution à find the level or R1,L that
makes V0 equal to the observed price. If V0 is too high we should use a higher rate à Goal seek function in excel. We will set
our value of V0 by changing the variable, R1,L and we will find V0. If V0 is below 100 it means that the rate that we used was
too high à worksheet C2.5 is the one that we are going to use to find the binomial tree that we are trying to calculate. We
should put V0=100.

Lecture 5

 Maturity strategy more attractive for lenders and more costly for borrowers
 The liquidity premium is positive
 The implied forward rate is the one that we extract from the observed spot rate

In the liquidity premium theory, if we consider the rollover strategy, we will renew at the expected future spot rate + the premium
that is required and the total return is higher than the maturity strategy.

The binomial tree is an arbitrage tree stochastic model of interest rates and we work with the lognormal model, developed by
Kalotay-Williams-Fabozzi:

 constant drift
 constant volatility
 log of the short rate is normally distributed

The arbitrage free models are calibrated based on their observed term structure of interest rates or on the prices of traded
bonds and the prices that these models are generating will be corresponding to the prices of the bonds on which they are
based.
We solve the quadratic equation with one unknown but it is only possible in a simple way: we calculate R1,L. We try to find the
level of R1,L which will make V0 equal to its price: 100

If we consider higher volatility, we will have lower rates on the downside and higher rates on the upside.

The bond without option is not function of the interest rate volatility: the value of the bond will be independent from the volatility
of the short rate. As soon as we introduce the embedded option, the higher the volatility the larger the value of the option.

Lecture 6
We saw how to calculate the binomial tree under the lognormal model assumption: it is an arbitrage-free model based on
observed on existing bond prices. We saw the case with 20% volatility of the short rate and the difference with the case where
the short rate volatility is 10%. The higher the volatility the wider the range of the rates for the different timesteps will be.

We extract the spot rates from the par yields via bootstrapping. The price of the bond (without embedded option) does not
depend on the volatility of the short rate
The lognormal model is used but the BDT model is more frequently used. The Black-Derman-Toy (BDT) model integrates mean
reversion for the drift term. When we have constant volatility in the Black-Derman-Toy (BDT) model, we get the same result than
in the lognormal model (KWF model).

In the BDT we have the stochastic differential equation that is used. Our drift term integrates two elements:

 The drift (function of time)

 Volatility of short rate

If sigma is negative, we have mean reversion. If the derivative is 0 we have the log-normal model and if sigma is positive we
have mean-fleeing. Flat vs. decreasing volatility term structure.

Log-normal process: no symmetry. If we use the BDT model with decreasing volatility or log-normal model with constant
volatility to value an option-free bond: we get the same price. As soon as we value an instrument that integrates an option to
interest rates, we will have a difference. Price sensitivity to interest rate changes: impact of using one of the two model is less of
a concern.

CHAPTER 3: systematic risk, duration, convexity, and key rate duration. Different tools to measure interest rate risk.

When we have option embedded in bond convexity is an interesting topic. We will see how convexity and duration are affected
when we have embedded option. Embedded option in bond: how to measure the spread on a bond over a reference rate (e.g.
treasury rate or the swap rate). Part of the spread that is measurable using the price is because we are purchasing/selling an
option. Option-adjusted spread is very important! Immunization is very important with rising interest rate. Duration to calculate
a hedge ratio to protect a bond portfolio in the short term against price risk. Key rate duration is linked to different segment of the
terms structure: we don’t have systematically parallel shift in the term structure. Link between duration and time à Immunization
strategy. Use of duration to calculate the portfolio YTM: the IRR of a bond portfolio. Systematic risk is non diversifiable risk and
is called interest rate risk. We have price risk and reinvestment risk. According to the investment horizon we have different
degrees of systematic risk. When you talk about systematic risk you need to know your investment horizon.

The duration is equal to the sum of the discounted cash flows, each cash flow being multiplied by its maturity, by the time up to
its maturity, until its maturity. Duration is a measure of effective longevity, it is a measure of maturity
Important the duration between two coupon payment dates. This time we calculate duration with the price. What is the price?
The full price! When you buy a bond on the coupon payment date: you pay the price as being equal to the discounted value of
the future cash flow but because the last coupon was already paid in there is no accrued interest that you have to pay. Between
2 coupon date you will have to pay to the seller the coupon that has accrued since the last coupon payment date.

Clean price = Full / dirty price – accrued interest

Duration is a measure of time and a weighted average of the time to maturity of the different cash flows/coupon and redemption
price and is therefore measured in years. The higher the coupon the smaller the duration compared to the final maturity of the
bond.

Duration is the weighted average of the different times to maturity of the cash flows of the bond. Sum of the present value of the
cash flows is the price of the bond.

Between two coupon date the duration declines at the same pace as time. At each coupon payment there will be a jump in the
duration. When we calculate the hedge ratios, we need to consider that there could be a jump in the duration either of the
portfolio that we are trying to hedge or in the hedging instrument that we are using. First derivative of the duration with respect to
time, when yields are unchanged is equal to -1.

Duration as a measure of price sensitivity to instantaneous changes in the yield. The duration is directly linked to the first
derivative of the price function of a bond to respect of its yield.

The duration is a price elasticity: it measures the relative price change due to a relative change in (1+y). The modified duration
is widely used by practitioner: Duration/(1+y).
If we know the modified duration, we have a first idea of the price change to which we will be exposed in case of a change
deltay in the yield.

With duration we are using a linear approximation to calculate the new price

Price-yield relationship suggested by using the duration of the bond.

 Yield increase: with duration you overestimate the price decline

 Yield decrease: underestimate the price increase

Underestimate or overestimate the new price with duration or modified duration: it is a linear approximation. The linear
approximation error will be larger the larger the yield change that we consider.

The first derivative of the discount rate with respect to y is negative. When y declines the duration increases. Duration calculated
at coupon date. What determines the maximum duration is not the time but the yield. A zero-coupon bond is called a pure
discount bond. A bond priced below par is called discount bond. For a bond priced below par the duration can go above the
limited represented by the perpetual bond but eventually will tend towards this limit.

Lecture 7

Using the duration, we neglect the convexity of the bond. Modified duration of the portfolio to measure the instantaneous price
sensitivity of a bond portfolio to a change in the yield. Dirty prices mean accrued interest included. True duration only in the case
the yield is the same for all bonds. The key rate duration: we use the sensitivity of a bond portfolio to certain segment of the
terms structure of interest rates. The duration and the modified duration give us a measure of the price sensitivity to
instantaneous changes in the yield.

We have a price sensitivity measure of instantaneous yield changes. Duration is not price volatility, is only a measure of price
sensitivity.

A greater price sensitivity is compatible with a lower price volatility. With a lower yield, a lower coupon and the same maturity,
ceteris paribus, we have more duration. Higher duration: more sensitive to instantaneous yield changes. To have price volatility
we need to have yield volatility.
The price appreciation is larger in absolute terms than price depreciation, this is convexity.

The duration is linked to the slope of the price function for a given price and a given yield. Convexity is positive for option-free
bonds. Convexity can be negative with callable bonds. When we have callable bonds and the yield is very high, the bond is not
a candidate for being called. A call option let the issuer call back its bond.

Lecture 8
The relative price change in absolute term will be larger than the relative price decrease in absolute term if the change
in the yield is positive: this is the convexity. Convexity: measures the rate of change of the slope of the price function.
Convexity is strictly positive for option free bonds. Convexity can be negative with callable bonds. When we consider lower level
of yields, there is a certain level of yields below which the call option starts to gain value and the value is increasing.

Convexity is directly linked to the second derivative of the price function. The price yield relationship is convex and non a linear
function. Taylor series: the relative price change is made up of 3 elements, the terms of first order and second order and what
we call the residue.

 The first derivative derived by the price is the modified duration

 The second derivative divided by the price is the convexity
 Residue: sum of higher order terms

When trying to measure the price sensitivity of the bond to an instantaneous change in the yield: we also consider the convexity.
The price-yield relationship is convex and not a linear function.

The dirty price is equal to the clean price, on the coupon payment date
Price approximation change with convexity.

With convexity and with increasing higher yield we will overestimate the price and underestimate the price change. In our Taylor
series the derivative of successive order is alternating in their signs. The derivative of odd order is negative, and the derivative
of even order are positive. If we have a yield decline, each additional term of our Taylor series is positive: when we have a yield
decline, we will each time get little bit closer towards the true price. This is not true when we are considering a yield increase.
The derivative of successive order is still alternating in terms of signs, but the different terms of our Taylor series are alternating.
Introducing the convexity will more than correct the underestimation that we have when considering only the duration. We are
overcorrecting the linear approximation error.

The limit is defined by the perpetual bond.

Convexity is desirable to have when interest rates are raising, we will lose less than what we gain in absolute terms compared to
situations where the yield increase. It has a limited impact, notably when considering option-free bonds. Convexity will be
positive for all option-free bonds.

Convexity can have a limited impact. Modified duration is way more important. Convexity is not paid for. Duration is
the main component of my price decline. The impact of convexity is limited compared to the impact of duration. To improve
convexity we reduce the coupon or increasing the maturity but then we will improve duration.
 Bonds with embedded options: we have early redemption clauses. This introduces uncertainty on the number of cash flows
and of the maturity of the cash flows and the uncertainty is linked to the level of interest rates.

Bond ranking is an important feature in the event of liquidation / bankruptcy of the issuer.

 Secured bonds: bonds that are secured (there is a collateral)

 Senior unsecured bonds
 Junior or subordinated debt

 When we buy a callable bond: we buy a non-callable bond + we sell the call option on the non-callable bond
 When we buy a puttable bond: we buy a non-puttable bond + we buy a put option on the underlying non puttable
bond

When the yield is equal to zero the price of the bond is equal to the sum of the contractual cash flows. For a callable bond, when
the yield is high the price of the call tends towards zero: there is no interest of the issuer to buy back its bonds. When we are
considering lower level of the yield, there is a level of Y* below which the call starts to have a value.

Lecture 9
In some cases, the price at which the bond can be redeemed early is at par. In many cases there is a premium on the
redemption price. The options are based on interest rates. Make-whole call option: the issuer has the right to redeem based on
the discounted value of all the remaining cash flow at any possible call date. The discounted cash flows being discounted based
on the reference treasury yield and a margin / spread that can go from 0 to 50bps, representing on average 15% of the yield
spread that the issuer must pay. There is a chance that the bond will be called not because interest rates go down but because
the spread / yield spread requested for the credit risk linked to the issuer is going down

There is another category of options: make-whole calls. Chart representing the value of the underlying call option with respect to
the yield level. This chart represents the value of a call option. As soon as we are using bond with embedded options the
formula of duration and convexity cannot be applied
With lower yields, we have negative duration that is increasing. More negative duration – less negative duration – more negative
duration: difficult to measure the price sensitivity of a bond with an embedded option, with regard to instantaneous changes in
the yield

How do we measure the duration of such bonds? The delta is the change in the call price divided by the change in the price of
the underlying

 For a call option, when we have very high levels of y we are out of the money and the delta is 0
 When delta tends towards 1, we should have a modified duration of the callable bond that tends towards zero
 Delta tends towards one when we are deep in the money or just in the money
 If the call date is tomorrow, very near. We have delta=1 when the option is ITM and the first possible call date is very near
(no convexity, no slope)

The gamma is the second derivative of the option price with regard to the price of the underlying: it measures the rate of change
of the delta. The gamma tends towards 0 both when the delta tends towards 0 and towards 1.

We use the binomial tree, and we are going to introduce a change in the par yield term structure.

We want to determine the duration, the modified duration, and the convexity for a bond with an embedded option (call/put
option). We apply a binomial / trinomial tree of interest rates to calculate the initial price for a given term structure of interest rate
and then to calculate P- and P+. convexity is negative with formula 3.22

Lecture 10
Formula 3.21 and 3.22 are very important, thanks to which we can calculate modified duration and convexity when we have
embedded options.
The binomial or trinomial tree are highly useful. They can make it possible to consider the features that are characterizing the
bonds with the embedded options. We can extract by bootstrapping the spot rate linked to the par yield term structure or
we can build the binomial tree with a lognormal model or with the BDT model. Whatever the arbitrage-free model that we
are going to use, as it is an option free bond, we get the same result.

At each node we won’t take the PV of future cash flows but we will consider that fact that at each node the bond can be called.
At each node where the exercise is possible, Vt will be the minimum of the strike price or the PV of the future cash flows. An
issuer will exercise if the PV of the future cash flows is larger than the strike price of the call.

The bond with the call will have a lower price than the price of the option free bond. As the volatility is larger, the value of
the call will be larger. Most of the callable bond are integrating a European style-like call option. We can only exercise it on a
given date (generally on an annual basis).

With higher rates, due to the fact that we have higher volatility, the value at each node will be lower (for a callable
bond). If we are considering a puttable bond, we are an investor that at each node should decide whether to exercise the option
to sell the bond at a predetermined price. The price of a puttable bond is the price of the underlying non puttable bond plus the
value of the put.

Puttable bond: if we have the PV > strike price, there is no exercise.

Higher volatility à higher put price. Floating rate notes are bonds where the coupon is based on the short-term rate. The issuer
sets a cap and offer a floor (nice for investor). A cap and floor, they are both options.
We are interested in the value of the option: this gives us a theoretical value of the callable bonds. We will recalculate the
binomial tree by introducing a shift in the flat term structure. We simulate the binomial tree. After calculating the modified
duration of the callable bond, we calculate the delta. With callable bonds, if we are ATM, we should have negative convexity .
The link between option value and volatility of the short-term interest rate.

What is happening to modified duration when we have a callable bond? With declining yields, the duration of the callable bond
increases, the MD increases and finally decreases until it joins the MD function of the non-callable bond and then it decreases
again.

The choice of the model matters when we are managing callable bonds. Practitioners are mostly using the BDT model: we
introduce non constant volatility. In terms of MD and convexity measures, the results are not so different.

Option-adjusted spread (OAS): We need to know what part of the total spread is due to the credit risk and what part is due to
the embedded option.

You express the value of the call option in terms of yield. In this case the 25 bps in terms of yield are paid because of the call
feature. Is the yield spread due to the risk that represents the issuer enough? A part is due to the option, which has nothing to
do with credit risk, and we don’t know if the credit risk is correctly paid or not. OAS is very useful.

To calculate the OAS, we need to use a binomial tree of interest rates. We use the lognormal binomial tree, and we are going to
calculate the spread that we must add to the risk-free rate calculated for each node that will make that the price that you get
corresponds to the price of the callable bonds. To know which part of the spread is attributable to the credit risk we add to the
reference rate at each node of the tree a spread, the same spread, the OAS . OAS: the credit spread, the spread requested by
the market for the credit risk linked to the issuer, and the remainder are due to the embedded call option feature.

 Log normal model: we assume that we have constant volatility

 BDT model: we introduce a non-constant volatility

Lecture 11 (excel workbook)

Priced at par: yield to maturity and coupon rate are the same.

The determinants of duration:

 Time to final maturity

 Coupon level
 Yield level
The lower the coupon the higher the duration, the lower the yield the higher the duration. Two bonds with the same modified
duration and a different convexity: if we expect an immediate increase of 60bps in the yield requested for the market for both
bonds: we should pick the bond with higher convexity. The choice would be the same in the case of a decrease in interest rates.
The choice will depend not only on convexity but also on the investment horizon, objectives and particularly and on the time
horizon.

Convexity and duration are part of the tools used to quantify instantaneous price sensitivity. When the yield for all maturities
tends toward zero the price of a bond (callable or non-callable) will tend to the sum of the contractual cash flows because the
discount factor tends towards 1.

OAS: the part that is due to credit risk. If we must invest for 2 days we shouldn’t invest in long dated bonds, exposing you to
relatively high price sensitivity. The OAS is the number of bps needed to be added on the one-year rate at each node of
the tree. If we add the spread at the interest rate at each node of the tree yields a price corresponding to the market price of the
callable bond.

Lecture 12 (workbook)
How options are impacting the price and how they affect the price sensitivity of the bond to changes in the yield.

If we want to calculate the OAS we have to enter the observed price for the risky callable bond.

 Callable bonds: we can have negative convexity

 Puttable bonds: we don’t have negative convexity

When do we have puttable bonds? MBS and convertible bonds.

We introduce section 5 now.

Immunization and how to protect a bond portfolio from subsequent changes in the yield. Determination of hedge ratios to hedge
ourselves in the short term against the change in the prices of bond that we have in our portfolio.

A yield to maturity is not a measure of total return (reinvestment risk). Duration is used to immunize the portfolio.

The element which has a positive impact on the total return has always a larger impact than the element which have a negative
effect or impact on total return. What happens to duration on the day of the first coupon payment date at unchanged yield: there
is a jump in the duration. The immunization condition is not satisfied anymore and then you need to reinvest the coupon in a
bond with a lower duration. on each coupon payment date, the condition that makes that the portfolio is immunized is not
satisfied anymore because of the jump in the duration.

Between wo coupon payment date all else unchanged the duration declines at the same pace as the passage of time.

If one week later the yield increase (portfolio immunized): we have a duration that will be higher but we don’t care. The
immunization not satisfied on two coupon payment date? Not a problem! We don’t have to adjust the portfolio for each change
in the duration introduced by changes in the yield. You need to consider adjusting on a coupon payment date.
Lecture 13

The major factor that affects bond returns is the shift factor, not the change in the slope, it’s not the change in the
convexity or in the curvature of the term structure of interest rates.

The portfolio duration must be equal to the time horizon, to have a portfolio that is immunized.

Hedge ratio: another application of the concept of duration, the difference is that we are specifically managing the price risk, we
are using the duration concept as a measure of the instantaneous price sensitivity to interest rate changes. We manage the
price risk in the short term. what are the main hedging instruments for bonds?

Lecture 14
Hedge ratio: main instruments to hedge a bond portfolio against price risk in the short term / near term . Specificity of
bond futures. If you are the seller of bond futures, you will choose the one to deliver: choosing the one that exhibits the lowest
net basis.

Conversion factor of the coupon-bearing bonds. The seller will choose the bond to deliver: the lowest net basis is going to be
delivered. Cash price vs. adjusted future price (future price x conversion factor of the considered bond ).

Net basis = subtracting from the gross basis the carry (= coupon income – financing costs to bear to buy forward). Lowest net
basis cheapest to deliver bond. Highest implied repo rate is linked to a combined transaction where you buy a deliverable bond
and sell the future and lend the bond on the lending market and on the delivery date of the future you will deliver the bond.

 Maximization of the yield that we are going to get in the transaction

 Choosing the bond that is the cheapest to deliver (lowest net basis or highest implied repo rate), if a seller
 Price of the future will reflect exactly the price fluctuations of the cheapest to deliver bond, through the conversion factor
The IRS is a tool to hedge the price risk in the short term: we can enter in an IRS of any maturity up to 50 years and we can
hedge in a much more granular and precise way the price risk that we bear in the bond portfolio. Dirty price= accrued interest
included. The goal in the first stage is to eliminate the price risk in the short term.

(3.28) formula for Hedge ratio in the general case when we have targets in the general case (target of modified duration). rates
going down: we add duration or swap with fixed interest floor (getting).

Lecture 15

Lecture 16
Recalculate for each term structure deformation. The price of bond for each deformation (relative price change)

We can write the price sensitivity profile of our bond int erm of partial duration.
Lecture 17

Lecture 18

Lecture 19
Lecture 20

Lecture 21
Lecture 22

Lecture 23
Lecture 24