Вы находитесь на странице: 1из 41

Chapter 6

Yield Curve Analysis II

c
@verview
M dhis chapter gets into more practical matters regarding
yield--curve analysis, although it also has some
yield
theoretical issues as well. We will examine:
M dhe differences between yield and price volatilities.
M How to calculate historical volatilities.
M Basic derm-
derm-Structure Issues
M Forward Rates of Interest
M Hypotheses of the derm Structure
M Bootstrapping and the extraction of zero coupon prices.

`
Yield and Price Volatility
M dhis is actually a pretty simple concept. Recall that modified
duration tells us (in percentage terms) the relationship between
price and yield.
M Basically as he shows in page 206, the price volatility is a
function of the yield volatility and duration:

0  º  º 


Yield and Price Volatility
M do calculate price volatility from historical data:
1. Compute the natural log of the price ratio:
Rt=ln(Pt+1/Pt) for each date t.
2. Compute the mean (  (  
   


. Compute the squared deviations for each t as xt=(Rt-
  
        

å  å 

    
 

   
 
 º 
å  å 

D
derm Structure Analysis
M *enerically, the phrase term-
term-structure refers to the
relationship between the maturity date of default-
default-free
zero coupon bonds and their yield to maturity.
M dhe yield on a zero coupon bond between time 0 and
time d (z=100/(1+yd)d) is called the spot rate of
interest.
M dhus, there is a spot rate from 0 to 1, 0 to 2, 0 to , etc.
M We also want to consider forward rates.
M We are going to jump ahead of the book and cover them
now, and then come back to extracting zeros.

Î
Spot and Forward Rates

M In normal, everyday usage, when we say the phrase


Ơinterest rateơ we are talking about a Ơspot rateơ.
M Somewhat formally, the n- n-period spot rate is the
interest rate charged on money borrowed at time 0
and repaid at the end of time n. We will denote this as
rn.
M Note that when rn is a zero coupon yield.

m
Spot and Forward Rates

*
   

  
 
 


  
!
 



# 



!"

Spot and Forward Rates
M A forward rate, however, is the interest rate associated
with a loan that you contract to today, but which will
occur at some future date.
M Note that once you sign the forward agreement, you are
bound to it, that loan will occur at the specified terms.
M We have to denote the beginning and ending points of
the forward rate. do do this we will use the notation fm,n
where m is the beginning date for the loan and n is the
ending date. You enter into the forward contract at time
0.
Spot and Forward Rates

*
   
 
  





"
    



# 

   



!

!"
6
Spot and Forward Rates
M @ur goal is to understand the fundamental relationship
between forward rates and spot rates, and how the
market enforces this relationship.
M Recognize that we define f0,1 M rm. dhat is, the initial
forward and future rates are the same.
M We also have to realize that there is no fundamental
difference between borrowing money on with a two
period spot rate, or by contracting to two consecutive
forward rates.

c
Spot and Forward Rates

 
  
$ % 
!  &  


'  


  $



 

 


!"

' 


(!   
$


 
!

 


!"

cc
Spot and Forward Rates
M Perhaps the easiest way to see this is with an example.
Letƞs say that you observe the following term structure
of interest rates:
Spot Rate Forward Rate
Rn fm,n

r1= 8 f0,,1= 8

r2= 9 f1,2 10.0009

r= 10 f2,= 12.0276

c`
Spot and Forward Rates
M do keep things simple, we will assume an annual
compounding frequency.
M Consider if you invested $1 at the two year spot rate. At
the end of the second year, you would have $1.188.


Ô )(!* )+

c
Spot and Forward Rates
M Similarly, if you were to lock in a series of two forward
rates, you would invest first at 8% for 1 year and then
at 10.0009 for the second year (but you lock in both
rates at time 0!)

Ô )(!+ (!!!* )+

cD
Spot and Forward Rates
M What if this were not the case? What if instead you
found a bank that were willing to loan to you one year
forward at 9%. How could you exploit this opportunity
for arbitrage?
M Clearly the bank is not charging enough in the second
year, so you want to borrow from the bank and lend to
the market.


Spot and Forward Rates
M You do this in the following way. First, you contract with
the bank at their (incorrect) forward rate of 9%.
M You then simultaneously borrow $1 in the spot market
for 1 year (at 8%) and lend in the spot market for 2
years at 9%.

cm
Spot and Forward Rates

„  


,*-


.

) 
$+-
   
$ .

+-


*-

!

,$


*-
 .

*-
$

 


!


Spot and Forward Rates
M dhe cash flows are relatively easy to work out:
M Borrow $1 at 8% for 1 year. At the end of the year
you must pay back 1*(1.08) = 1.08.
M You must, therefore borrow $1.08 from the bank at
9% for the second year. You will have to pay back a
total of $1.08*(1.09) = $1.177.
M You invested $1 at time 1 for 2 years at 9%. You will
receive $1*(1.09)2 = $1.188 when that loan matures.

c
Spot and Forward Rates
M We can thus look at the payments on all legs of this
trade ƛ in particular note the Ơnetơ position:

drade dime 0 dime 1 dime 2

Borrow $1 at 8% +1.00 -1.08


for 1 year
Lend $1 at 9% -1.00 +1.188
for 2 years
Borrow (through +1.08 -1.177
forward) $1.08
for 1 year at time
1 at 9%.
å    

c6
Spot and Forward Rates
M Literally there is no risk to you, and you are guaranteed
an positive cash flow later ƛ hence this is an arbitrage ƛ
as the old Dire Straights song goes Ơmoney for nothingơ.
M Clearly arbitragers would quickly take advantage of this
mispricing in the market, and Ơdisciplineơ the bank for
this.

`
Spot and Forward Rates
M do see a little more complicated example, letƞs look at a
real--world case.
real
M dhis is taken from the Wall Street Journal of January 22,
2000.

`c
Spot and Forward Rates
M @n January 22, we observed the following zero coupon
bond prices:
M Zero maturing in February, 2011: $58.6875
M Zero maturing in February, 2012: $55.125
M dhe yields on these bonds are:
M $58.6875 = 100/(1+r20 /2)20: r20 = 5.40
M $55.125 = 100/(1+r22/2)22: r22 = 5.488

``
Spot and Forward Rates
M dhe one year spot rate beginning in year 10, i.e. in
period 20 (since we are now back to semi-
semi-annual
compounding) is given by:

ƒ  !++  
    
 
  /0
!
  !!  !

    

 

`
Spot and Forward Rates
M Letƞs say that you now found a bank that was offering
forward rates of 9% for year 10. How could you exploit
this?
M You would clearly want to borrow at the market rate of
6.47% and lend to the bank at their incorrect rate of
9%.

`D
Spot and Forward Rates
M dhe trade you would put together would be as follows:
M Borrow $1000 in the spot market for 11 years at 5.488%
M Lend $1000 in the spot market for 10 years at 5.40%.
M Use the forward rate to lock in to lend to the bank at time 10 for
1 year at a rate of 9%.


Spot and Forward Rates
M dhe cashflows would be:
M Note that if you lend $1000 at 5.40 for 10 years, at the end of the
10 years you would receive 1000*(1+.054/2)20 = $1,70.76.
M You would then lend this amount to the bank at 9% for the 11th
year. At the end of the 11th year the bank would pay back to you
170.76*(1+.09/2)2= 1,860.55
M You would then have to pay back the original $1000 you borrowed
at 5.488%, which would be: 1000*(1+.05488/2)22 = $1,814.02.

`m
Spot and Forward Rates
M We wind up at the end of year 11, then, receiving
$1860.55 and paying back to the market $1814.02 ƛ a
net gain of $46.52
M So just like before we can see that we earn Ơmoney for
nothing.ơ


Spot and Forward Rates
M We can thus look at the payments on all legs of this
trade ƛ in particular note the Ơnetơ position:
drade dime 0 dime 10 dime 11

Lend $1000 at -1000 +170.76


5.4% for 10 years

Borrow $1000 for +1000 1814.02


11 years at
5.488%

Lend to bank -170.76 +1860.55


through forward
at 9%

å    

`
dheories of the derm Structure
M @ver the years various researchers have attempted to
relate spot rates to forward rates ƛ that is one period
future spot rates to currently observed forward rates.
M Four primary hypotheses have emerged:
1. dhe Expectations Hypothesis
2. Liquidity--Premium Hypothesis
Liquidity
. Market Segmentation Hypothesis
4. Local--Expectations Hypothesis
Local
M Letƞs briefly examine each one.

`6
Expectations Hypothesis
M dhere are multiple variants of this hypothesis. dhey
ultimately all try to develop the notion that forward rates
are some form of the marketƞs estimate of future spot
rates.
M dhe unbiased expectations hypothesis formally states
just that, i.e. fk,k+1= Et[R*k].
M dhe difficulty is, that empirically forward rates are
terrible predictors of futures spot rates. If these are the
marketƞs expectations, the market consistently sets its
expectations of future forward rates too high.


Liquidity--Preference dheory
Liquidity
M dhis says that lenders would, all things equal, prefer to
lend for shorter periods of time. do induce them to lend
long term, therefore, borrowers must pay an additional
premium, which shows up in the form of an upward
sloping yield curve, which generates forward rates that
are greater than the marketƞs actual expected spot rate,
i.e. fk,k+1= Et[R*k] + IJk,k+1.
M dhe difficulty is that sometimes the yield curve is
inverted, which would imply that investors preferences
changed.
M dhis is, in some ways, the most convincing of these
arguments.

c
Market--Segmentation
Market
M dhis is not so much a theory as an anti-
anti-theory: basically
it says that the long-
long-term and short-
short-term markets are
different, and that only if they are grossly out of line
with each other will participants in one market cross
over to participate in the other.
M Basically it says that the fundamental premise of a yield
and/or forward rate coupon curve is flawed: that the
curve tends to give the illusion that there is a
relationship that is not there.
M *enerally this theory is not particularly useful, especially
in the context of fixed income pricing.

`
Local Expectations Hypothesis
M dhis is a special case of the expectations hypothesis. It
basically says that all bonds (of the same credit quality)
have the same rate of return for a very short period of
time. dhis is the only arbitrage-
arbitrage-free model!
M ƠShortơ is normally defined to be whatever is the
smallest time period in the model being used, i.e. its
instantaneous for continuous time models, but maybe
one month for discrete-
discrete-time models.
M Some of the discrete time models really push this to the
absolute limit.
M Formally it is: E[Zd-(t+1)]/Zd-t= (1+rt)


Coupons, Zeros and Strips (oh my!)
M In the market we primarily observe coupon bearing
bonds, although many times we wish to work with zero
coupon bonds ƛ this is especially true in the context of
building arbitrage-
arbitrage-free term structure models.
M We really have two ways of getting zero information:
extracting them from coupon bearing bonds in a process
known as bootstrapping
bootstrapping,, and directly observing them in
the strips market.
M Why ever use the bootstrapping method? Primarily
because the coupon market is still larger and more liquid
than the strips market, and as a result traders have
more confidence in the quotes from that market.

D
Bootstrapping
M Bootstrapping is simply a procedure for extracting zero
coupon bond rates from coupon bearing bonds. It is
most commonly used in the dreasuries market.
M dhe basic idea is that you start with an initial short term
bond ƛ typically a 1 month or  month bond, which is
truly a zero coupon bond, and you calculate its yield.
M You continue calculating the zero coupon bonds until you
run into your first coupon-
coupon-bearing bond. For an
exampleƞs sake, letƞs say that you observe a six month
and a twelve month zero (which is typical), and that
your first coupon bearing bond matures at month 18.
Denote the zero coupon yields as Z6 and Z12.


Bootstrapping
M dypically you use a par
par--value coupon bearing bond, that means one
for which you know the price is 100. *iven that you know the values
of Z6 and Z12, what you do is find the value of Z18 that makes this
statement true:
1 1 !! M  1 
!!  M 
M "
ó M / 1  ó M  1  ó M + 1 

M For example, let us say that Z6=10%, Z12 = 12%, and that the
coupon on a par-
par-priced coupon bond is 1%. dhe 18 month zero
would be:
" 1  " 1  !!  " 1 
!!  
ó! ó!/ ó  l+ 1  "



!/
++! "
2 3+ "
ó  l+ m
Bootstrapping
M @ne major problem with bootstrapping is that you donƞt
really get bonds spaced evenly every six months, and if
you are working with a model where you need monthly
zeros, you really have trouble finding it.
M Constant Maturity dreasuries (CMd) help this a lot. dhey
provide you on a daily basis with rates for bonds that
have maturities of 1 year, 2 year,  years, 5 years, etc.
M You still have to make assumptions about what happens
in-
in-between those bonds. @ne approach is to get more
data ƛ i.e. find interest rate derivatives that do mature
between the bonds.


Bootstrapping
M dhe second approach is to make some assumption about
what is going on in-
in-between the observed rates. In
essence you simply assume how the rates will behave.
M A common (but not particularly good) assumption is to
assume that the par bonds are linear in coupon between
observed points. dhat is, if Cn and Cn+2 are the coupons
of par bonds maturity at times n and n+2, and you need
a coupon for a bond which matures at time n+1 but
there is no such bond trading, simply use linear
interpolation to estimate Cn+1, i.e. Cn+1 = (Cn+Cn+2)/2
M dhe problem with this approach is that it can lead to
some rather bizarre-
bizarre-looking forward rate structures.


Bootstrapping
M @ther commonly used methods include various curve
fitting algorithms such as piecewise cubic splines, cubic
Ơdiscount rateơ generating functions, and statistical
methods such as Diamentƞs method outlined in
Sundaresanƞs book.
M While cubic-
cubic-spline methods are pretty common in
commercial software, there is still a lot of variation in
how various firms elect to fit their curves. All methods
involve tradeoffs, such as being willing to accept a more
Ơjaggedơ curve in some regions in exchange for a
smoother curve in other regions.
M dhere is not one universally accepted method.

6
A Simple Cubic
M @ne method that is sometimes used is to assume that
there is a specific functional form for the discount
function, d(t).
M Realize that the discount function is just that ƛ it tells
you how much a dollar to be received at time t is worth
today ƛ and from that you can obviously calculate the
zero coupon rate.
M A potential method is to estimate the parameters of this
formula using regression analysis:
d(t) = 1 + at + bt2 + ct

D
A Simple Cubic
M dhen, when you need the zero coupon rate for  time
t, you simply plug in the value of t, and it spits out the
discount price, and you can then solve for the zero
coupon yield.
M Unfortunately, there is no guarantee that it will exactly
fit the current yield curve, and there are number of yield
curves that we do occasionally see, such as inverted
curves, which is it very difficult or impossible for the
simple cubic function to match.
M A cubic spline really just takes this a step further. Please
see the handout for an example.

Dc

Вам также может понравиться