Lecture 2 InterestRates Presentation2
Lecture 2 InterestRates Presentation2
Felix Matthys
2 Bond Pricing
3 Yield To Maturity
6 Accrued Interest
Explanations:
• Having money today gives you the opportunity to invest it and earn a risk free rate or
consuming goods and services.
• Furthermore, the purchasing power of your money can decline if price level rises in the
future.
• Want criteria to choose today from two different future payments, in other words we are
interested in transforming the future payments, which may be deterministic or random
into a value today.
• This is done through a discount factor Z . We have
PV = FV × Z
• Similarly, given a cash flow of payments with values C0 , ..., Cm , where Ci is the
payment at the end of the ith period, the present value of the cash flow is
(compounded every period):
T
C1 C2 Ct X Ct
PV (0) = C0 + + 2
+ ... + t
=
(1 + r ) (1 + r ) (1 + r ) t=1
(1 + r )t
• It can be checked that the present value PV and the future value FV (T ) are
related by:
FV (T )
PV =
(1 + r )t
Example: On Monday, the 12th of January 2016 (= t) the discount factor was
Z (t, T ) = 0.999825 (mid quote) with payment of $ 1 on February 15th (= T ) which implies a
(annual) interest rate of 0.1936%.1
No Arbitrage Condition
For any time t ≥ 0, and let T1 < T2 and assume that the nominal interest rate r be positive,
then we must have that
Z (t, T2 ) ≤ Z (t, T1 ), ∀t ∈ [0, T1 ]
1
To compute this interest rate, we use expression (2), and invert it to obtain the
continuously compounded interest rate r (t, T ) = − τ1 ln (Z (t, T )), where τ = 33/365.
Felix Matthys Fixed Income ITAM 6 / 90
Compounding Frequency
for r (t, T ).
Continuous (n = ∞) 110.52
Felix Matthys Fixed Income ITAM 9 / 90
Relationship between Discount Factors & Interest Rates
Conversion Formulas
Given the interest rn (t, T ) with n times compounding frequency, the continuously
compounded interest rate r (t, T ) satisfies
1
e −r (t,T )(T −t) = Z (t, T ) = n(T −t) (3)
rn (t,T )
1+ n
0
1990 2000 2010
Figure: Panel A: Yield curve. Source: U.S. Department of the Treasury. Panel B: Yield Curve
evolution over time. Source: Federal Reserve Board.
Pz (t, T ) = N × Z (t, T )
where N is the face, principal, amount or par value which usually is $100.
Note: Subscript z refers to ”zero coupon bond”.
where Z (t, Ti ) denotes the discount factor for each payment date i and
m = ⌊(T − t) × n⌋ is the total number of coupon payments over the lifetime of the
bond.b
a
We will use the convention that the last payment date Tm is just T , which is equal to the maturity of the bond
b
Note: The operator ⌊·⌋ simply rounds any number down to its closest integer. Example: ⌊2.4⌋ = 2 or ⌊2.9⌋ = 2
Observe: Equation (4) shows that the price of a coupon bearing bond can be expressed
as a sum of zero-coupon bonds.
and that coupons are paid semi-annually, i.e. n = 2. Why is this an arbitrage?
• At time t, buy bond Pc (t, T ) and sell c/2 × N units of zero coupon bonds with
maturities Ti , i = 1, 2, . . . , m − 1 and (1 + c/2) units of a zero coupon bond with
maturity Tn = T .
• Payoff: ∆ = nc m
P
i=1 Pz (t, Ti ) + Pz (t, T ) − Pc (t, T ) > 0
• For every coupon payment date Ti , the arbitrageur obtains a coupon c/2 and has
to deliver exactly −c/2 treasuries to the investor. Thus the trader is fully hedged.
An important observation
• The bond pricing Equation (4) reveals that, a coupon paying bond, can be
interpreted as a portfolios of zero coupon bonds (discount factors), i.e.
securities with just a single payment at maturity.
• Thus, once we have obtained the zero coupon bonds, we can price any other
security.
• Furthermore, given a coupon paying bond, we can exactly replicate its payoff
profile by taking positions in a set of zero coupon bonds.
• Since the two securities, i.e. the coupon bond and the portfolio of zero coupon
bonds have identical payoff profiles (payment amounts are the same and they
occur at the same time), their price, by the law of one price, has to be the same.
Otherwise there exists an arbitrage opportunity.
• To replicate the 43 s of November 30, 2011, we use the 1 21 s due Nov. 30, 2010, the
4 87 s due May 31,2011, and the 4 21 s due Nov. 30, 2011
• Number these bonds 1, 2, and 3, and let θi be the face amount of bond i in the
replacing portfolio.
• The following equations show the requirement that the cash flows of the
replicating portfolio equal those of the 34 s on each of the three cash flow dates.
Replicating portfolios are easier to describe and manipulate using matrix algebra
1.25% 4.875% 4.5% 1 0.75%
1+ 2 2 2 θ 2
0 1+ 4.875% 4.5% θ2 = 0.75%
2 2 2
0 0 1+ 4.5%
2
θ3 1 + 0.75%
2
which can be easily solved by pre-multiplying each side by the inverse of the cash flow
matrix.
Felix Matthys Fixed Income ITAM 20 / 90
Determining Replicating Portfolio Weights
We have a system of three equations and three unknowns. Thus there exists a unique
solution to the portfolio holdings θi , i = 1, 2, 3
• The equations can be solved one-at-a-time, since one bond matures on each date
θ1 = −1.779%
θ2 = −1.790%
θ3 = 98.166%
What are the costs associated with setting up this arbitrage opportunity?
1 Bond 1: −100.55 × 0.01779 = −1.789
2 Bond 2: −104.513 × 0.0179 = −1.871
3 Bond 3: 0.98166 × 105.856 = 103.915
Therefore the total costs of the replicating portfolio are 100.255 which is higher than the
price of bond 4 (100.190). Since they both deliver the same payoff for each of the three
maturities, there exists an arbitrage opportunity.
• An arbitrageur could buy 100 face amount of the 3/4% coupon bond for $100.19
and sell the replicating portfolio for $ 100.255 looking in an immediate profit of
0.065 per unit or 6.5 cents.
• If the arbitrageur scales up the trade to $ 1 Billion face value, the riskless profit is
$ 650’000.
(i) coupon 1 41 s 4 78 s 4 21 s 3s
4
(ii) maturity 30 nov. 2010 31 may. 2011 30 nov. 2011 Portfolio 30 nov. 2011
The previous example has shown that we can price a coupon bond using two approaches
1 Pricing of the coupon bond by deriving the set of discount factors Z (t, T )
2 Find a replicating portfolio, consisting of a set of spanning bonds, that matches
the payoff characteristics of the bond we want to price
Remark:
• In real-world markets with transaction costs, taxes etc, these results only hold
approximately (prices must be within ranges)
C×θ =c (5)
where θ ∈ Rn denotes the vector of portfolio weights and c ∈ Rn denotes the cash flows
of the bond to be replicated.
′
1 Then the price of the bond we want replicate using discount factors is c(t) × Z(t).
2 The price of the same bond using the replication strategy can be expressed as
′
P(t) × θ
Thus, if the market is free of arbitrage, then we must have that
′ ′
c(t) × Z(t) = P(t) × θ
′ ′
c (t) × C−1 × P(t) = P(t) × C−1 × c(t) (6)
Since both expressions in Equation (6) are scalars, this proves the claim.
Felix Matthys Fixed Income ITAM 25 / 90
Yield to Maturity (Bond Yield)
Important question: How can we measure the expected return on a bond investment?
• Consider a zero coupon bond with time - to- maturity τ = T − t. Assuming the
bond is held until maturity τ = 0 the annualized return is2
" 1 #
N n×τ
yz := yz (t, T ) = n × −1 , τ ≥0 (7)
Pz (t, T )
• For a bond that pays a coupon rate c% per year, it is no longer that obvious how
to measure the return on investment.
⇒ For general investments it is called the internal rate of return, IRR. The IRR of a
bond is its yield or YTM.
2
Compare this expression to Equation (3).
Felix Matthys Fixed Income ITAM 26 / 90
Yield to Maturity (Bond Yield) cont.
• Suppose that a bond pays a face value equal to N and n coupon payments per
year at Ti , for i = 1, . . . , m, each of size c/n, at regular intervals.
• Suppose there are T − t years left until maturity for a total of ⌊(T − t) × n⌋ = m
periods until maturity.
• We can see that the yield yn is the interest rate implied by the current market price
of the bond (IRR).
Felix Matthys Fixed Income ITAM 27 / 90
Comparing Bonds by YTM
Suppose you have the following two bonds available:
Bond 1 Bond 2
Maturity 2 years
Bond 1 Bond 2
Maturity 2 years
Proposition:
Let a, b ∈ N and q ∈ R, then we have the following result
b
X q a − q b+1
qt =
t=a
1−q
Proposition:
Let a, b ∈ N and q ∈ R, then we have the following result
b
X q a − q b+1
qt =
t=a
1−q
Proof:
Remarks:
• When c = yn , the bond trades a par value, i.e. Pc (t, T ) = N.
• When c > yn , the bond trades at a premium to par, Pc (t, T ) > N.
• When c < yn , the bond trades at a discount to par, Pc (t, T ) < N.
Remark:
• The par rate above can not be computed using the yield to maturity y , as the yield
to maturity is the one interest rate that gives us the current bond price Pc (t, T ).
• If the yield to maturity is equal to the par rate, then we must have that the bond
price trades at par, i.e. Pc (t, T ) = N.
Questions:
• What is the market price of the bond at time t?
• Is the bond selling at par, at discount or at a premium? Is c > y or the reverse?
• What is the par rate c2,p for this bond?
• What is the continuously compounded par rate cp ?
• Suppose today t = 0 and that we have available the price of bonds for every
maturity Tj = 0.5, 1, 1.5, ....Tn . Inverting (8), we obtain
P1 (0, 0.5)
Z (0, 0.5) =
N (1 + c1 /2)
P2 (0, 1) − Nc2 /2 × Z (0, 0.5)
Z (0, 1) =
N (1 + c2 /2)
P3 (0, 1.5) − Nc3 /2 × (Z (0, 0.5) + Z (0, 1))
Z (0, 1.5) =
N (1 + c3 /2)
• In general, for every i: P
i−1
Pi (0, Ti ) − Nci /2 × j=1 Z (0, Tj )
Z (0, Ti ) =
N (1 + ci /2)
Felix Matthys Fixed Income ITAM 37 / 90
Bootstrapping the zero bond curve
Exercise (Bootstrapping)
Suppose we have the following data:
Exercise (Questions:)
• Given a face value N = 100$, what are the bootstrapped discount factors Z (t, T )
for each maturity Ti , i = 1 . . . , 7?
• Determine the zero coupon yield curve, i.e. compute r2 (0, i/2), for every
i = 1 . . . , 7.
• Essentially, each row i of C corresponds to the cash flows of bond i during all the
maturities T1 ,...,Tm . In contrast, each column j describes all the cash flows that
occur on that particular maturity Tj from the m bonds.
• So, the discount function can be obtained simply by
Note: We need the cash flow matrix C to be invertible for this approach to work. In
other words, we need exactly the same number of bonds n and maturities Ti ,
i = 1, . . . , m.
Felix Matthys Fixed Income ITAM 44 / 90
Regression Approach
• The main problem with extracting zero coupon bonds using the matrix algebra is
that we rarely have nicely spaced data.
• Sometimes we have too many maturities and sometimes we do not have any
maturity.
• The regression methodology deal with the case where there are too many bonds
compared to the number of maturities.
• This is typically the case when we consider maturities up to 10 years.
• For example, there are 81 quoted coupon bonds with maturity less than 5 years.
• But there are only 60 months in 5 years, implying that for this horizon we have too
many bonds.
• where the data are yi = Pi and xij = Cij , and the regressors are βj = Z (t, Tj );
• From basics OLS, we then find that
b (t) = C′ × C −1 C′ × P (t)
Z
Main idea:
• Assume a parametric form of the discount factor Z (t, T ), that is take an explicit
functional form for Z (t, T ) and then try to match it to the data (current time t
bond prices). The discount factor is given by
Advantage:
• This approach gives us a continuous function in time to maturity for the yield
r (t, T ). In other words, for every time to maturity τ ≥ 0 we can compute the
corresponding yield r (t, T )
Problem with this approach:
• We need to estimate/calibrate parameters (estimation error). What is the right
optimization technique?
• The functional form we use for the yield r (t, T ) might be inappropriate for
modeling the term structure. For some types of curves, the assumed parametric
form of the yield r (t, T ) might not be able to reproduce the market data
(especially when yield curve is highly non-linear).
• Early suggestion due to Nelson and Siegel: Assume Z (0, T ), is such that:
Z (t, Tj ) = e −r (t,Tj )×(Tj −t)
(Tj −t)
1 − e− λ (Tj −t)
r (t, Tj ) = θ0 + (θ1 + θ2 ) (Tj −t)
− θ2 e − λ
where Θ = {λ, θ0 , θ1 , θ2 }.
• Put differently, we try to make the difference between the model implied prices
M D
Pc,i (t, Ti ) and their market (data) prices Pc,i as small as possible.
Felix Matthys Fixed Income ITAM 50 / 90
Curve Fitting
• The Nelson Siegel Model works nice, but it lacks the flexibility to match term
structures that are highly non-linear.
• Lars Svensson proposed a simple extension to the model, which is the one most
widely adopted.
• In particular, we assume:
• Market conventions often cause confusion in the study of fixed income securities.
• Conventions vary according to the location and the type of instruments one is
concerned with.
• Two instruments that are similar may be quoted in very different ways depending
on where the are traded.
• In financial markets, we always have two prices. The ask and the bid price. The
ask price is the price at which the market is willing to sell. The bid price is the
price at which the market is willing to buy.
• The difference is called the bid-ask or bid-offer spread.
• A typical quote for deposits in London money and foreign exchange markets is
Ask Bid
5 14 5 81
30 + 30 + 12 = 72
days, while the actual calendar would give 73 days. More interestingly, the same
instrument purchased on February 28, 2003, and sold the next day, on March 1,
2003, would earn interest for 3 days. Yet, a money market instrument such as the
interbank deposit would have earned interest only 1 day.
Days 30
• Depositor earns interest on the trade date, but not for the day the contract
matures.
• We are looking at the deal from the bank’s side, since we quote the offer rate.
• The interest rate earned is expressed using the formula r · ∆(Days) × 100, 000.
• The full (invoice,cash) and quoted price are also known as the dirty and clean
prices, respectively.
• The clean price is typically the quoted price on financial news sites.3
3
This is the case for the United states, whereas, for instance in Europe, the quoted price is usually the dirty price.
Felix Matthys Fixed Income ITAM 58 / 90
Example: Accrued Interest
a
Note: The day count convention is actual/actual.
Felix Matthys Fixed Income ITAM 62 / 90
Felix Matthys Fixed Income ITAM 63 / 90
Felix Matthys Fixed Income ITAM 64 / 90
Floating Rate Bonds
where N is the face value, rn (t, t + 1/n) is the n month Treasury rate at time t and s is
the spread.
• If n = 1 annual coupon payments and r1 (t, t + 1) denotes the annual treasury rate
at time t, if n = 2 semi-annual coupon payments and r1 (t, t + 1/2) denotes the
semi-annual treasury rate at time t, etc.
• Convention: Each coupon date is called the reset date, as it is the date when the
new coupon is fixed according to Equation (66).
• Thus, only the next coupon is known at the current date t. In other words,
r (t, t + 1/n) is known at t but not r (t + 1/n, t + 2/n) and r (t + 2/n, t + 3/n) etc.
are all unknown at time t!
Felix Matthys Fixed Income ITAM 66 / 90
Floating Rate Bonds cont.
Graphical illustration:
Consider a one year floating rate bond with semi-annual coupon payments and the
spread s is equal to zero.
• The coupon payment at time t = 0.5 depends on today’s interest rate r2 (0, 0.5)
which is known.
• Suppose r2 (0, 0.5) = 2% then c(0.5) = 100 × 0.02/2 = 1.
• The coupon at time t + 1 will depend on the prevailing interest rate at t + 0.5
until t + 1, i.e. r2 (0.5, 1) which we do not know today at t = 0.
• Thus the cash flow at t + 1 is 100 + c(1) = 100 × r2 (0.5, 1)/2
• Question: How do we price this bond when we do not know the size of the
coupon payments at t + 1, i.e. the coupon rate is random?
1 + r2 (0, 1/2)/2
FRB(0) = 100 × = 100
1 + r2 (0, 1/2)/2
• Therefore,
• Lagging the coupon payments by precisely six months leads the cash flow and the
discount effect to exactly cancel each other out and therefore leaving the bond
value at par at any reset date.
If the spread is nonzero, we can decompose the coupon payments into a two parts.
1 The floating rate which depends on the reference interest rate
2 and the fixed spread s .
In other words,
Floating Coupon with spread s = Floating Coupon with zero spread + Fixed coupon s
Thus the price of a floating rate bond with spread s denoted by FRBs (t) with n coupon
payments per year and T years until maturity, decomposes as follows,
m
s X
FRBs (t) = FRB(t) + N × Z (t, t + i/n), m =T ·n
n i=1
m
s X
= 100 + N × Z (t, t + i/n)
n i=1
Suppose that T1 , T2 , . . . , Tn are the floating rate reset dates per year with T − t years
until maturity. Let the current date t be between (Ti , Ti+1 ). Then the price of a
semi-annual floating rate bond with spread s is given by
s
FRB(t, T ) = N × Z (t, Ti+1 ) (1 + r2 (t, t + 1/n)/n) + ,
n
where Z (t, Ti+1 ) is the discount factor from t + u to Ti+1 and m = n(T − t). At the
reset dates, which corresponds to t = Ti in Equation (75), we obtain that
1
Z (Ti , Ti+1 ) = r2 (Ti ,Ti+1 )
1+ n
and thus
m
s X
FRB(t, T ) = 100 + ×N × Z (t, t + 1/n · i)
n i=1
• Given price P (t, T ) and cash flows c/2 at Ti ’s for i = 1, .., n, and 1 + c/2 at Tn ,
the (semi-annually compounded) yield-to-maturity YTM is defined as that rate y
such that
n
X c/2 100
P (t, Tn ) = + (11)
y 2×(Ti −t) y 2×(Tn −t)
i=1 1 + 2 1+ 2
• Recall that given the spot curve r2 (t, Ti ), we also had
n
X c/2 100
P (t, Tn ) = +
i=1 (1 + r2 (t, Ti ) /2)2×(Ti −t) (1 + r2 (t, Tn ) /2)2×(Tn −t)
1 2 9% 98.57 9.82%
3 2 3% 88.00 9.91%
N1 × 9 + N2 × 10 = 3
N1 × 109 + N2 × 110 = 103
• Solving, we find that a portfolio long N1 = 7 units in the first bond and short
of N2 = −6 units in the second bond mimics exactly the cash flows of bond
3.
• The value of this portfolio is N1 × 98.57 + N2 × 100.34 = 87.95 < 88.
• The YTM of the portfolio is even bigger: YTM=9.94%.
Table: Yield Curve on March 15, 2000 calculated from CRSP data (Daily Treasuries).