10/21/2016
Intro to Interest Rate Modeling
Fixed Income Securities S10
Prof. Jijo Lukose P.J.; IIM Kozhikode
Topics
Differences in modeling interest rates and equities.
Dynamic modeling of term-structure movements
ensuring no-arbitrage.
Comparison of equilibrium and no-arbitrage models of
the term-structure of interest rates.
Introduction to Interest rate models
Discrete Models
Continuous-time Models
Term Structure models
Fixed Income Securities S10
10/21/2016
Differences in Modeling Interest-Rates vs
Equity Modeling
The following three features of equity processes are not true of
interest-rate processes:
The price of the underlying security follows a geometric Brownian
motion.
The interest rate is known and constant.
The volatility of the returns is constant.
Interest rates are not lognormal. Nor are bonds, as they revert to
par whereas lognormal securities do not have an expected value of
par at T.
Assuming constant interest rates for options on interest rates is a
contradiction in terms. Bond prices cannot be stochastic if interest
rates are constant.
Return volatility of bonds varies over time. For example: when
approaching maturity, the bond is reverting to par and volatility
dampens.
Fixed Income Securities S10
Modeling Interest-Rates
It is easy to model the underlying drivers of bond pricesthe interest rate processes.
But, while modelling interest rates, we have to model the
entire yield curve.
Therefore it is important t ensure that models are
internally consistent and free of arbitrage.
Fixed Income Securities S10
10/21/2016
Arbitrage Violations Example
It is not as easy to make sure that there is no-arbitrage
on the tree. Here is an example of what might happen if
you are not careful.
Undertake a strategy to buy the two-year zero at time
zero, and sell it after one year. The funding cost is 10%.
Fixed Income Securities S10
Arbitrage Violations Example
Arbitrarily moving the term structure up and down often leads to
arbitrage. Here, we see that even a simple parallel shift of the term
structure is fraught with error.
Borrowing cot is 10%. Therefore the amount to be repaid is
81.16x1.1=89.276.
Thus the model permits arbitrage
Fixed Income Securities S10
10/21/2016
Term Structure Models
Pricing of interest rat derivatives requires understanding of
evolution (& volatility) of future interest rates
correlations between future interest rates
The best is to model the dynamics of the entire term structure/yield curve
Modelling the entire yield curve is practically not feasible. Modelers
typically assume there are only a small number of factors that influence all
interest rates. These factors are referred to as the models state variables.
Empirical Facts
Interest rates for various maturities tend to move in the same direction.
Historically the term structure of volatility of interest rate is an increasing
function until the one-year maturity and a decreasing function for longer
maturities-humped form;
there seems to exist some correlation between interest-rate volatility and
interest rate level
Interest rates generally exhibit some mean reversion.
Fixed Income Securities S10
An Example
YTM on ZCB 1 year =10%, 2 year =11%, 3 year=12%.
Begin with an interest rate tree for the evolution of spot
rates:
Arbitrage-free (risk-neutral) probabilities
and prices for the first period.
p1=0.55
1p1=0.45
Fixed Income Securities S10
10/21/2016
Using the three-year zero
We need to find Bu and Bd
We also need to find the
probability p2
Final Short-Rate Tree
p2
Valuing Options
Once we derive the tree with interest
rates at each node and RNP on each
branch, we can price interest rate
derivatives.
9
Fixed Income Securities S10
Limitations
Though the above model ensures that the bond prices
converge to par at maturity, it still lead to extremely high
interest rates in future. This happens due to the fact that
probability of an up move or a down move is
independent of the level of interest rates.
Modeling trees with mean-reverting interest rates.
Set an upper limit 2 and lower limit 0.
Probabilities are dependent on the current interest rate
p=1-rt/ 2 and 1-p=rt/ 2
Fixed Income Securities S10
10
10/21/2016
No-Arbitrage vs Equilibrium Models
No-arbitrage models work through the following steps:
Identify portfolios that are equivalent.
Impose the condition that their costs should be the same.
Derive the consequences of this condition for security prices.
Examples:Vasicek (1977); Richard (1978); Brennan & Schwartz (1979).
Equilibrium models
These modelsderive term-structure dynamics in a general
equilibrium setting.
Cox, Ingersoll, and Ross (1985) proposed a different approach to
term-structure modeling that was based on constructing a full
dynamic general equilibrium model in which the prices of all assets in
the model were simultaneously determined.
Since a set of prices cannot represent an equilibrium if arbitrage is
possible, their approach guaranteed the absence of arbitrage (that is,
it was sufficient for no-arbitrage).
Fixed Income Securities S10
11
A Comparison
Equilibrium is sufficient for no-arbitrage.
No-arbitrage models (the first was Ho-Lee 1986) exactly
fit the initial yield curve, equilibrium models do not.
Equilibrium models have fewer free parameters hence
cannot match all yields.
Additional time-varying risk premium parameters may be
added to equilibrium models to match existing yields
exactly.
Fixed Income Securities S10
12
10/21/2016
Equilibrium models Vasicek Model
dr ( r )dt dz
Where is the long - run mean of the short term rate
(kappa) is the speed of adjustment of the short - term rate to
the long run mean
2 is the variance parameter
The deterministic component goes to zero when interest rate is equal to
mean. In this case, interest rates follow a random walk.
Fixed Income Securities S10
13
CIR Model
CIR Model
dr ( r )dt r dz
The variance of the changes in interest rate is dependent
on the level of rates.
Fixed Income Securities S10
14
10/21/2016
Arbitrage-Free Models of the yield curve
BDT
Ho and Lee
Hull and White
HJM
Fixed Income Securities S10
15
One-Factor Models
Pros
Highly tractable and parsimonious
Models allow closed-form solutions for bonds and options
Cons
Cannot fit the observed/actual yield curve
Important implications: all bonds are perfectly correlated
Market says otherwise
Vasicek allows for negative interest rates
Possible shapes not very rich: monotonically increasing or
decreasing with slight humps
Fixed Income Securities S10
16
10/21/2016
Overview of the BDT Model
The Black-Derman-Toy (henceforth, BDT) model has as its objective the
construction of a model of interest rate evolution that satisfies two
important properties:
The model is arbitrage-free.
The model is consistent with the current term-structure of interest rates and
volatilities.
Main Assumptions
To achieve these ends, BDT posit a single-factor, discrete-time model of the
interest rate process.
All movements in the yield curve are derived in an arbitrage-free manner from
movements in the short rate.
The short rate process itself is assumed to follow a lognormal distribution on a
Binomial tree.
It is further assumed that the volatility of the short rate depends only on time.
It is assumed throughout that the risk-neutral probabilities of up and down
moves are each 1/2.
Single factor model => All bond price movements are perfectly correlated over each
period.
Lognormal short rate => Negative interest rates are precluded.
Fixed Income Securities S10
17
BDT Procedure
The evolution of short rates in the binomial tree is
determined using two inputs:
The current term structure of interest rates.
The volatilities of yields of different maturities.
Bootstrapping is employed:
Attractive features:
Simplicity.
Positive interest rates.
Consistent with given term-structure of rates and volatilities.
Fixed Income Securities S10
18
10/21/2016
HJM and LMM
HJM models describe the arbitrage-free movement of the
entire forward curve. LMMs describe the movement of
forward Libor.
Drifts of the forward rates are functions of volatilities
identification of risk premiums not required.
Both frameworks admit multiple factors.
Lognormal versions of these models results in closedform, Black-Scholes type option formulae.
Both models are easy to calibrate to the interest rate
markets, but LMMs allow direct calibration.
Fixed Income Securities S10
19
10
