Learning Module 2

The Arbitrage-Free Valuation Framework

Excellence Demands Great Education 52

Agenda

Learning Module 2: The Arbitrage-Free Valuation Framework

• Explain the arbitrage-free valuation of a fixed-income instrument
• Calculate the arbitrage-free value of option-free, fixed-rate coupon bond
• Describe a binomial interest rate tree framework
• Describe the backward induction valuation methodology and calculate the
value of a fixed-income instrument given its cash flow at each node
• Describe the process of calibrating a binomial interest rate tree to match a
specific term structure
• Compare pricing using the zero-coupon yield curve with pricing using an
arbitrage-free binomial lattice.
• Describe path-wise valuation using binomial interest rate tree and calculate
the value of a fixed-income instrument given its cash flows.
• Describe a Monte Carlo forward-rate simulation and its application.
• Describe term structure models and how they are used

Excellence Demands Great Education 53

 Arbitrage Free Valuation Framework
• Arbitrage free valuation is the single key concept used when valuing
bonds, derivatives and financial assets.

• Key features to arbitrage: 1) Quick and executable strategy

2) Riskless profit 3) No cash outlay required.

There are two types of arbitrage opportunities:

1. Value addition: when the value of an entire investment differs from the
sum of its parts.
2. Dominance: a financial asset with a risk-free payoff in the future must
have a positive price today i.e. that two securities that have risk free
payoffs at the same time t must have the same discount rate 𝑟 .

• Otherwise, there exists an opportunity to make money with zero risk.

• Arbitrageurs exploiting such opportunities (buying the cheap or constituent
securities and selling the expensive or the full investment) will force market
prices to converge until no further opportunities exist.
Excellence Demands Great Education 54
 Example

Q1: Consider two bonds:

Bond X trading for $100 Payoff in 1 year = $105
Bond Y trading for $200 Payoff in 1 year = $220
Bond X and Y are similar in terms of risk and cash flow stream. Does an
arbitrage opportunity exist? Describe how a trader would exploit the arbitrage.

Q2: Which of the following investment alternatives includes an arbitrage

opportunity?
Bond S: The yield for a 3% coupon, 10 year annual pay bond is 2.5% in NYC.
The same bond sells for $104.376 in Chicago.
Bond T: The yield for a 3% coupon, 10 year annual pay bond is 3.2% in Hong
Kong. The same bond sells for RMB97.220 in Shanghai.

Excellence Demands Great Education 55

 Valuing Option-Free Bonds

Recall: The traditional method for determining bond price involves the
following Three steps :
1) Estimate the cash flows: Coupon and Principal
2) Determine the appropriate discount rate
3) Calculate the present value of the estimated cash flows.
 This approach involves using a single discount rate (YTM) to determine
bond price. Implicit is the assumption that the yield curve is flat!!

Arbitrage Free Valuation Approach

 However, a correct approach to valuing bonds is to view a coupon bond as
a package (or portfolio) of several zero-coupon bonds. With each zero
coupon bond’s cash flow to be discounted at the respective spot rate.
• If the theoretical (calculated) price > actual (observed) price. The bond is
trading cheap. Buy bond (whole), strip the bond and sell it as zeros (parts)
• If the theoretical price < actual price. The bond is trading expensive. Buy the
individual zero’s (parts) reconstitute the bond, sell it as a full bond (whole).
Excellence Demands Great Education 56
 Example
Sara Jane, CFA has just completed bootstrapping the treasury yield curve, she
provides you with the following spot rates:
Year Spot Rates
1 3.8%
2 4.0%
3 4.2%
4 4.3%
5 4.7

Calculate the value of a 5 year, 6 % annual pay Eurobond with $100 face value.
Also calculate the YTM on the bond.

Assume the Eurobond is trading in the market at par. What arbitrage trades
shall you execute?
Excellence Demands Great Education 57
 Valuing a Bond with Embedded Options

• The bond valuation method used for the previous example is only apt for
option free-bonds and bonds with zero cash flow uncertainty. The method
cannot be extended to value bonds with embedded options because:

Complication 1: bonds with options have future cash flows that depend on
future interest rates (eg. callable, putable bonds)

Complication 2: future interest rates are volatile.

• To aid our valuation process we:

o Assume a constant interest rate volatility and
o Use an interest rate ‘tree’ (also called lattice model) to model future
interest rates paths and predict bond cash flows along each path.

• We can use the interest rate tree to value both, bonds that are option-free as
well as bonds with embedded options.

Excellence Demands Great Education 58

 Binomial Model
Model features:
1) Read the tree from left to right. Start with the initial rate (left) and move
towards the right through a number of possible increases/decreases in rates.
2) Initial rate is the discount rate for cash flows occurring at the end of period 1
3) Each period is called a ‘time step’ and is of equal length.
4) The binomial tree is a discrete-time model (aka, lattice model)
5) Assumes a constant interest rate volatility (σ).
Period 3

Period 2 𝑖
Period 1 1/2 𝑖
𝑖 𝑖 Interest rate at each node
Period 0 1/2
is the 1 period forward
Initial rate 1/2 𝑖 𝒊𝟐 interest rate.
𝑖 𝑖 Eg. uuui is the 1 period
1/2 𝑖 forward rate 3 years
1/2 𝑖 𝑖 from now.

Excellence Demands Great Education 59

 Binomial Model
Model features:
6) From one period to the next, interest rates can realize one of two possible
rates, that is, either up (U) or down (D).
7) The probability of interest rates moving up or down is exactly half.
8) The path the interest rates can take is displayed as an interest rate tree.
9) To determine the interest rates on the tree we need:
• Interest rate model – that will predict a series of interest rates overtime with
randomness. Eg. lognormal random walk.
• A volatility assumption (historical volatility or implied volatility)
10) The use of the random walk model provides two desirable properties:
• Higher volatility of interest rates at higher rates.
• Non-negative interest rates.
11) As we move from one time step to the next, the volatility of interest rates is
assumed to increase by 1 standard deviation each time.
12) From the interest rate model, determine 𝑖 . Then calculate:
i i e
i i e
Excellence Demands Great Education 60
 Binomial Model
Model features:
13) From the interest rate model, determine 𝑖 (one period forward rate, two
periods from now). Then calculate:
i i e i i e
14) If you have i , you can directly obtain i by using the relationship:
i i 𝑒

Excellence Demands Great Education 61

 Example
Q1: Assume a lognormal random walk process is used to determine interest
rates on a binomial tree. If 𝑖 0.98% and the volatility of interest rates is
assumed to be 𝜎 15%, calculate the interest rates on the following nodes:
a i
b i

Q2: If the lowest possible interest rate at time period 3 is given as 𝑖 , what
is the mathematical relationship of this lowest rate to 𝑖 , 𝑖 and 𝑖 ?

Excellence Demands Great Education 62

 Binomial Model
Constructing an Arbitrage Free Interest Rate Tree
• Interest rate tree is constructed using computer software.
• It is based on the fundamental principle – use the tree to price an on-the-run
issue, the price should be arbitrage free, i.e., bond value = market price.
• For a tree with ‘n’ periods there will be 2 unique paths, where 𝑛 1
• Adjacent forward rates are two standard deviations apart i.e. 𝑒
• The middle rate on the tree for a period is ~ the one period forward rate

Option Free Bond Valuation using Binomial Model

• Bond valuation using an interest tree applies the ‘backward induction’
methodology.
• Backward induction is a process of valuing a bond from reverse, that is,
starting from maturity and moving backwards to time t = 0
• To determine the value of a bond at T = T – 1:
− Start from T = T (maturity),
− Calculate the average of the bond cash flows using risk neutral probabilities (0.5)
− Discount the proceeds back one period to T = t – 1.
− Appropriate discount rate is the forward rate associated with node under analysis

Excellence Demands Great Education 63

 Example

Given the interest rate tree shown below, what is the value of a 5% coupon, 3
year option-free Treasury bond?

𝑓 = 5.343%

𝑓 = 4.625%

𝑓 = 4.375%
𝑟 = 4.0%

𝑓 = 3.787%

𝑓 = 3.582%

Excellence Demands Great Education 64

 Path-wise Bond Valuation
• An alternate approach to pricing a bond i.e. path-wise bond valuation is to
list out all the unique interest rate paths a bond can be valued at
• Then price the bonds using each of the unique interest rate paths. Note: the
interest rates are forward rates; compute the geometric average of interest
rates (spot rates) before discounting cash flows.
• Lastly, compute an arithmetic average of all bond prices to arrive at the final
bond price.
Recall: The bond is a 5% coupon, 3 year bond

t=0 t=1 t=2 Bond Price

4% 4.625% 5.343%
4% 4.625% 4.375%

4% 3.787% 4.375%
4% 3.787% 3.582%

Average Price

Excellence Demands Great Education 65

 Monte Carlo Forward Rate Simulation
• A key assumption of valuing bonds using the binomial tree is that: ‘cash
flows on the bond are independent of the path that interest rates take’.
• The above is true for default-free, option-free bullet maturity bond.
• The assumption is unrealistic for bonds with embedded options, such as
residential mortgages, callable bonds, putable bonds etc.

• For RMBS securities, the cash flows depend not only on the level interest
rate at a point in time (lower rates, higher prepayments) but also on the path
(re-financing burnout) that interest rates take over time
 This implies that RMBS and similar securities have cash flows that are path
dependent.
 Due to this path dependency, the simplistic backward induction approach is
replaced with a more complex Monte Carlo approach to valuing bonds.

Monte Carlo Approach

• Simulation randomly generates ‘000s of interest rate paths consistent with a
volatility assumption and an assumed interest rate probability distribution

Excellence Demands Great Education 66

 Monte Carlo Forward Rate Simulation

Monte Carlo Approach

• Each simulated interest rate path is ‘calibrated’ i.e. used to value a actively
traded benchmark security to ensure the calculated price ≅ market price.
• If the interest rate path produces a security price higher (lower) than market
price, a constant rate ‘C’ is added (deducted) to all the rates along a path until
calibrated.  Drift adjusted calibration model.
• For each interest rate path, cash flows are calculated for each period using an
assumed prepayment speed, default rate, recovery rates etc.
• The cash flows along each path are discounted using the respective forward
rates to arrive at a bond price.
• The final bond price is an average of the prices determined across all paths.

Excellence Demands Great Education 67

 Term Structure Models

Aim: to model interest rate movements over time.

Challenge: to model the real world phenomena of moving rates, restrictive
assumptions are made.

• Term Structure Models mathematically describe how interest rates evolve.

• No one model perfectly captures the interest rate dynamics.
• Models are based on a set of simplifying assumptions. Despite the simplicity
the models are useful for pricing and hedging.  Modelers face the trade-off
between simplicity and accuracy when selecting a term structure model.

• Term Structure Models can be ‘fit’ to binomial models i.e. determine the
values of the interest rates on the up node / down node etc.

• Modelling future path of rates is not only critical for scenario analysis &
stress testing of bonds, but also important in the valuation of complex fixed
income instruments (e.g. MBS, callable, putable bonds)
Excellence Demands Great Education 68
 Term Structure Models

• Term Structure Models vary based on 3 key characteristics:

1. Interest Rate Factors: One factor models vs. Multi factor models
o One factor models (short rate, or one-period rate) are the simplest
kind of models; they assume that a single short period rate drives the
movement in the entire term structure i.e. all rates move in the same
direction (e.g. parallel moves), although they don’t have to
necessarily move by the same amounts (e.g. steepening/flattening).
o Multifactor models incorporate additional factors such as slope
(twists in yield curve) and curvature (bowing in yield curve)

2. Interest Rate Process: Models use a stochastic (jumpy) process to

describe rates. Model has 2 components: 1) drift term, 2) stochastic term
dr θ dt σ dZ
Change in rate drift change in time volatility random term

• Although the models are based on continuous time, the models can
still be ‘fit’ to discrete period binomial trees.
Excellence Demands Great Education 69
 Term Structure Models

• Term Structure Models vary based on 3 key characteristics:

2. Interest Rate Process:
• Drift term 𝜃 𝑑𝑡 describes the expected (zero-volatility) rate path. The
drift term can be set to be either ‘constant’ OR ‘mean-reverting’.
• Stochastic term 𝜎 𝑑𝑍, adds randomness or volatility to the rates. The
term Z is a Weiner process that is normally distributed  Since
normal distribution extends from +∞ to − ∞, it is possible for such
models to produce interest rate paths with negative rates.

3. Class of Model
A. Equilibrium Models: model considers fundamental economic
variables that influence rates & uses them to model changes in rates.
• Unlike arbitrage free models, equilibrium models do not take the
yield curve as ‘given’ and do not force model parameters to take
on values that will produce a model bond price the same as current
market price. Neither do the models force the modelled yield curve
to equal the currently observed yield curve.
Excellence Demands Great Education 70
 Term Structure Models

• Term Structure Models vary based on 3 key characteristics:

3. Class of Model
A. Equilibrium Models:
• Practitioners prefer equilibrium models since they capture not just
the current market environment as reflected in the term structure
but also the possibility of many different future paths.
• Compared to arbitrage free models, equilibrium models do not
require a lot of parameters to be estimated. This allows for cost
saving & efficiency, but this may reduce model accuracy.
• Equilibrium models are best used for relative bond valuation.
• Two famous models used: 1) Cox-Ingersoll-Ross, 2) Vasicek
• Both CIR and Vasieck model are single factor which assume that a
single observable short-term rate 𝑟 drives changes in all
interest rates i.e. parallel shifts in yield curves.
• Multifactor models are superior to single factor models, but are
complex and cost time and money to implement.
Excellence Demands Great Education 71
 Term Structure Models

• Term Structure Models vary based on 3 key characteristics:

3. Class of Model
B. Arbitrage free (no-arbitrage) models:
• Model takes the yield curve as ‘given’ i.e. assumes currently
traded bond prices are correct. Model then uses observed bond
prices to arrive at (parameterize) possible movement in rates.
• “Parameterizing” is the process of determining the values of
variables in the model such that those parameters produce bond
prices that match current market prices  Large number of
parameters must be estimated, this increases the computational
effort which is undesirable.
• A key difference is that equilibrium models use real probabilities,
whereas arbitrage-free models use risk-neutral probabilities.
• Models are widely used in practice and are favoured by users since
they give bond prices that = market price. A useful property for
pricing bond for trade & pricing hedge instrument for current time
• Two models frequently used are: 1) Ho-Lee Model, 2) Kalotay-
Williams-Fabozzi (KWF) Model
Excellence Demands Great Education 72
 Equilibrium Term Structure Models

Cox-Ingersoll-Ross Model
• Assumes interest rates are market equilibriums determined by the supply of
savings and demand for investment funds.
• The model is: 𝑑𝑟 𝑘 𝜃 𝑟 𝑑𝑡 𝜎 𝑟 𝑑𝑧
Drift Randomness
Where:
𝑑𝑟 = small change in short term interest rates σ = interest rate volatility
𝑑𝑧 = small random walk movement 𝑑𝑡 = small change in time
k = speed of mean reversion (higher ‘k’, faster mean reversion)
𝜃 = long run short-term interest rate r = the short term interest rate
𝜃 𝐫 = distance of rate from the mean

• The model has two parts: 1) a deterministic part (drift) and a

2) stochastic part (randomness)
• The drift forces the interest rates to move towards its long-run rate ‘𝜃’ at a
speed determined by parameter ‘k’
Excellence Demands Great Education 73
 Equilibrium Term Structure Models

Cox-Ingersoll-Ross Model
• Model’s random element suggests that interest rate volatility depends on the
level of interest rates i.e. high (low) interest rates  higher (lower) volatility.
• The random element (𝜎 𝑟 𝑑𝑧) is assumed to follow a standard normal
distribution with mean zero and standard deviation of 1.

Example:
Narnia a country in western Antarctica, currently has short-term interest rates of
3% with a long-run value for short term rates observed to be at 8%. Assuming a
speed of adjustment of 0.4 and annual volatility of 20%. Using the CIR model:
• Advise in which direction would next period’s short term interest rate be
headed (i.e. higher or lower)?
• Estimate the short term interest rate over two periods where the random walk
change 𝑑𝑧 is 0.5 in the first period and -0.1 in the second period.

Excellence Demands Great Education 74

 Equilibrium Term Structure Models

Vasicek model
• Vasicek has the same drift term like CIR i.e. Vasicek model assumes that
short-term interest rates will reverts to its mean long-run value.
𝑑𝑟 𝑎 𝑏 𝑟 𝑑𝑡 𝜎𝑑𝑧

• Like CIR, in the Vasicek model the volatility term follows a random normal
distribution for which the mean is zero and the standard deviation is 1
• Key difference in Vasicek model is that the randomness of interest rates is
constant i.e. independent of the level of interest rate  Implying that low
rates and high rates both have equal volatility.
• A key feature of Vasicek model is that it does not force interest rates to be
non-negative i.e. the model allows for negative interest rates.

 Note: both CIR and Vasicek model may result in a term structure of interest
rates that does not match the observed yield curve.
 However, if analyst believes parameters of the models (a, b, σ) are correct
then he/she can use the estimated rates to identify mis-pricings in the market.
Excellence Demands Great Education 75
 Arbitrage Free Models
• The ability to calibrate models to match market data is a desirable feature of
this model and highlights the key drawback of CIR and Vasicek models.
• CIR and Vasicek models have only a small number of free parameters (k, 𝜃
and σ) that need to be estimated. Whereas, arbitrage free models allow
parameters to vary deterministically with time, requiring a large number of
parameters to be estimated and thus allowing for market yield curve to be
estimated with accuracy.
• Since the yield curve is modelled with accuracy, it can be used to value
derivatives and bonds with embedded options.

Ho-Lee Model 𝑑𝑟 𝜃 𝑑𝑡 𝜎𝑑𝑍

• Ho – Lee model is constructed (calibrated) using observed market prices to

find the time dependent drift term that generates the current term structure.
• Calibration is performed using a binomial tree, where the time dependent
drift parameter 𝜃 changes from period to period i.e. for each time step.
• Like Vasicek, Ho-Lee model has a constant volatility, and this implies that
the model allows interest rates to become negative.
Excellence Demands Great Education 76
 Arbitrage Free Models

Kalotay-Williams-Fabozzi Model 𝒅 𝐥𝐧 𝑟 𝜃 𝑑𝑡 𝜎𝑑𝑍

• Like the Ho-Lee Model, the KWF model assumes a time-dependent drift, no
mean reversion and constant volatility.
• Key difference is that KWF models describes the dynamics of the log of the
short rate not the dynamics of the short rate.
• As a result, the KWF model assumes that the log of the short rate is
distributed normally, implying that short rate itself is distributed lognormally.
• Translation: KWF models the change in short rate as normally distributed,
and the rate as log-normally distributed
• Modeling the log of the short rate will prevent negative rates.
• A key issue with KWF model is that the tails of the log-normal distribution
are thinner than real world distributions, and this has pricing implications for
interest rate options.

Excellence Demands Great Education 77

 Multi-Factor Models

• All models (CIR, Vasicek, Ho-Lee, KWF) discussed thus far were single
factor models. An example of the multi-factor model is Gauss+
• The Gauss+ model is used extensively in valuation and hedging.
• Gauss+ model incorporates short-, medium- and long-term rates:
o Long-term factor is mean reverting and reflects trends in macroeconomic
variables.
o Medium-term rate also reverts to the long-run rate.
o Short-term rate does not exhibit a random component, which is consistent
with the central bank controlling the short end of the rate curve.
• The above assumptions results in a hump-shaped volatility curve across
tenors, with medium-term rates being the most volatile.

Excellence Demands Great Education 78

 Exam Style Questions
Q1: A $1,000 par, annual-pay Treasury bond matures in 3 years and has a
coupon rate of 9%.
Maturity Yield on Maturity Strip
The value of this Treasury
bond is closest to: 1 year 8%

A. $1,001.46 2 year 8.5%

B. $986.62 3 year 9%

C. $1,000.58

Q2: The table below shows a bond traded on three exchanges. Based on the
bond prices, and ignoring transaction costs, the best action that an investor
should take to profit from the arbitrage opportunity is to:

Eurex NYSE Euronext Frankfurt

Price €103.7956 €103.7815 €103.7565

A. Buy on Frankfurt, sell on Eurex

B. Buy on NYSE Euronext, sell on Eurex
C. Buy on Frankfurt, sell on NYSE Euronext
Excellence Demands Great Education 79
 Exam Style Questions
Q3: Using the binomial tree, the arbitrage-free value of a 3-year 6% coupon
bond is closest to:
𝑓 = 7.0053%

𝑓 = 5.4289%

𝑓 = 5.7354%
𝑟 = 3.5%

𝑓 = 4.44448%

𝑓 = 4.6958%

A. $102.42
B. $101.85
C. $103.58

Excellence Demands Great Education 80

 Exam Style Questions
Q4: An analyst has provided you with an incomplete interest rate tree:
Current Year 1 Year 2 Year 3 Year 4
1.2500% 1.8229% 1.8280% 2.6241% Node 4 – 1
1.4925% Node 2 – 2 Node 3 – 2 4.2009%
1.2254% 1.7590% 3.4394%
𝜎 10% Node 3 – 4 2.8159%
Node 4 – 5
Which of the following statements about the missing data is correct?
A. Node 3 – 2 can be derived from Node 2 – 2.
B. Node 4 – 1 should be equal to Node 4 – 5 multiplied by 𝑒 .
C. Node 2 – 2 approximates the implied one-year forward rate two years from
now.

Excellence Demands Great Education 81