Relative Value Models (Feb04)

Bond Relative Value Models
Detecting relative value refers to the process of comparing the potential returns
of alternative investments. This paper describes practical Excel/VBA
implementations of two such relative value models which provide information on
over-, respectively underpriced bonds. The first one is a simple model based on
traditional yield to maturity analysis (explained in section 1) and the second one
demonstrates the slightly more complex JP Morgan discount function model, an
approach of term-structure of interest modelling (shown in section 2). Both have
been tested in practical applications in bond relative value research but they
have here been adapted for academic purposes. In particular, the real time data
feeds have been generalized to be independent of a particular data provider. To
use them in a “real world” setting, they would require links to current price and
bond data as well as utility macros to maintain the bond baskets analysed.
The Excel files containing the models can be downloaded from the following
website:
http://www.mngt.waikato.ac.nz/kurt/
The detailed link is currently

http://www.mngt.waikato.ac.nz/kurt/frontpage/ModelMainpages/FixedIncomeMo
dels.htm#BondRelativeValueModels
February 2004
Kurt Hess
Senior Fellow
Department of Finance
Waikato Management School, University of Waikato,
Hamilton, New Zealand
[email protected]
____________________________________________________________________________________________
Kurt Hess, [email protected] 23/02/2004

Relative Value Models (Feb 23 04).doc
Page 2 of 18
1 Simple yield to maturity based benchmarking

Yield to maturity, also called redemption yield based measures to find relative value in a
universe of bonds is the traditional method used by most practitioners in the field.
As is commonly known, yield to maturity assumes that an investor holds the bond to
maturity and all the bond’s cash flow is reinvested at the computed yield to maturity. It is
found by solving for the interest rate that will equate the current price to all cash flows
from the bond to maturity. In this sense, it is the same as the internal rate of return (IRR)
defined in many finance textbooks (e.g. in Reilly & Brown, 2002). Some technical
complications arise from the treatment of accrued interest which according to most
market conventions have to be paid upfront by the bond buyer. These and other issues
related to bond yields are discussed in specialized resources such as Fabozzi (1999)
The yield to maturity based model shown here has two sub-components, each stored in
a separate Excel workbook.
• ConfBenchmark (Feb04).xls shows the implementation of a redemption yield

based benchmark model for the universe of Swiss government bonds. It is
explained in section 1.1.
• Cheap Rich List (Feb04).xls produces a cheap/rich analysis for a universe of

Swiss domestic corporate bonds using the benchmark curve generated in the
Confbenchmark-model. It is explained in section 1.2.
1.1 Benchmark curve

The graphical depiction of the relationship between the yield of the bonds of the same
credit quality but different maturities is known as the yield curve. If one was able to
subdivide the universe of bonds into homogeneous baskets of bonds, drawing a yield
curve for each of them would allow benchmarking within the group. Unfortunately this is
not a simple undertaking as will be discussed in the next section. What is generally
done, is to compare yields is a well-established liquid benchmark curve. For most
markets this would be the government bond market but other rates such as swap rates
are also used a benchmarks.
Figure 1: Generic Time / Yield Chart with Bond Yield Curve
4.0% Yi
3.5%
Yi *
3.0%
2.5%
Yield
2.0%
1.5%
1.0% Yield Benchmark Bonds
0.5% Benchmark Curve
0.0%
0 2 4 6 8 10 12
Years to Maturity
____________________________________________________________________________________________

Page 3 of 18
There are a number of methods to derive such a yield curve. For a small universe of
benchmark bonds, as for instance the one of New Zealand government bonds, one
could simply “link” the few bonds in the time/yield chart with straight line connectors
which means any intermediate yield would be found through linear interpolation. The
model implementation shown here derives a benchmark yield curve from observed
bond redemption yields of Swiss government bonds by fitting a polynomial that
minimizes the squared error of redemption yields. Many other ways such as fitting cubic
splines are employed in financial modelling software. This is not to speak of numerous
more advanced methods for yield curve interpolation and fitting for which, as an
example, the introduction to Krippner (2002) lists the leading references.
The following paragraph derives the calculation of the polynomial coefficients.
Defining …
Redemption yield bond i: Yi with i = 1,2, … to n (number of bonds in basket)
Interpolated yield bond i: Y *i = am t im + am−1t im−1 + ... + a1t i + a0
where
ti : time to maturity of bond i
a0 , a1 ,..., am : coefficients of m degree polynomial
n
∂ ∑ (Y *i −Yi )
2
n
… we need to minimize ∑ (Y *
i =1
i −Yi ) ⇒
2 i =1
∂ak
= 0 for k = 0,1,2 up to the
desired dimension of the polynomial m. These are m+1 equations for the unknown
coefficients a0 , a1 ,..., a m .
Some algebra shows that a0 , a1 ,..., a m must be a solution of the following system of
linear equations:
 n

∑t n
i ∑t n
i
2
.. .. ∑t
n
i

m
  a0   n
 ∑ Yi 

 ∑ ti
 n
∑t n
i
2
∑t n
3
i .. .. ∑t
n
i
m +1
  
  a1   n
 ∑ i i 
Y t
 ∑ ti2 ∑t 3
∑t 4
.. . ∑t m+2 
 a2   ∑ Yiti2 
 n n
i
n
i
n
i
×  =  n 
 : : : . :  :  : 
     
 : : . . :  :  : 
 tm
∑
n
i ∑t
n
i
m +1
∑t
n
i
m+2
.. .. ∑n ti2m  am  ∑n Yitim 
This system is typically solved numerically by one of many well known algorithms such
as the ones described in Press et al. (1992). Yet because a coefficient like
∑n
ti2 m becomes an extremely large number for higher values of m, algorithms will lose
accuracy. However, this is not an issue in most cases because meaningful

interpolations will not exceed 3rd to 4th order polynomials. The VBA implementation in
the model uses the so-called LU decomposition to solve the system as described in
Press et. al (1992) p.43.
____________________________________________________________________________________________

Page 4 of 18
1.2 Redemption yield based cheap rich analysis

Corporate bonds are not homogeneous in their credit quality so bond analysts compare
their yield to some well established benchmark, e.g. government bond curve. This yield
differential is then measured in basis points (bps) and gives an indication of the credit
risk priced into the corporate bond.
An example: a five year government bonds yields 4.5%, a corresponding corporate

bond 5.2%. The yield differential or credit spread is 0.7% which equals 70 bps.
This section first discusses the nature of these credit spreads, followed by a description
how these have been modelled here. It concludes with a simplified numerical example
of cheap-rich analysis.
1.2.1 The nature of credit spreads

What should be the proper credit spread for a particular bond?
Practitioners typically compare a particular credit spread of with the ones mapped for
bonds of comparable credit quality. Many bonds now have credit ratings by one of the
leading agencies so they might base this comparison on averages observed for each
credit quality. Deviations from these averages are then used to generate a cheap/rich
lists of corporate bonds. This simple method corresponds to a parallel shift of the
benchmark yield curve to account for the credit risk of a particular rating category.
Beyond this, there is obviously the whole body of credit risk modelling literature that
tries to develop approaches to pricing claims subject to credit risk. A result of such
modelling is often a so-called “term structure of credit spreads” which plots the credit
spread as a function of time to maturity of the risky bond.
The small chart to the right

plots such a term structure Figure 2: Credit Spreads: Typical Shape
derived by a model of Credit Spread (in bps)
Longstaff & Schwartz (1995)
which is explained in more 300
detail in Appendix 1. What is 250
more important here is that it 200
150
reflects a peculiarity of credit
100
spreads often observed in
50
the markets. They are not -
constant as assumed in the 0 2 4 6 8 10 12
simple approach above but Time to maturity
for short-term bonds the

spreads become tiny because the likelihood of events pushing weaker credits into
bankruptcy becomes minute. There is also the typical “hump” which indicates higher
spreads for medium term than for long-term bonds. The hump becomes very
pronounced for lower non-investment grade quality (e.g B-rated bond) since bonds are
then no longer traded on a yield but on break-up value basis. On the other hand, it can
hardly be observed for high-grade bonds.
____________________________________________________________________________________________

Page 5 of 18
1.2.2 Defining the term structure of credit spreads in the model
The model presented here lets the user choose the desired term structure of credit
spreads for each rating category by means of shape parameters. This is illustrated
through a numerical example in Figure 3 below. The term structure is basically broken
down into two sub-periods. A short- to medium term period to T∞ (T_indef) which is
followed by the long-term characteristics of the credit spread.
Figure 3: Explanation Shape Parameters Term Structure of Credit Spreads
250
300bps/yr (Slope at t=0)
Term Structure for a4 = 3
200 T∞ Term Structure for a4 = -3
Credit Spread
T_indef=3
150 -12 bps/yr (Slope at
t=T_indef)
S_indef=125
100 Spread Lower Limit

(70.0% of 125bps)
50 Short-term characteristics long-term

of credit spreads characteristics
6.125
0
0 1 2 3 4 5 6 7 8
Years to Maturity
As to the first period, a fourth order polynomial is fitted between zero and T∞.
Spread (t ) = a 0 + a1t + a 2 t 2 + a 3t 3 + a 4 t 4
It must meet the following three boundary conditions:

• must be zero a time = 0 which means a 0 = 0
• must reach S∞(S_indef) at time t=T∞.
• slope at point S∞/T∞ must equal the slope of the long-term curve beyond T∞.
(more details on this slope are below)
The user affects the shape of the polynomial with two parameters.
• The initial slope (in bps per year) at time = zero may be specified
• The shape of the hump is also affected by parameter a4 which corresponds to
the fourth order coefficient of the polynomial. Figure 3 shows the term structure
for a4= 3, respectively minus 3.
In the long-term horizon, the user can specify a slope for the further development of
credit spreads. In most cases it will be set to or close to zero to obtain constant credit
spreads beyond time T∞. There is also the option to specify a lower limit below which
credit spread may never decline. This parameter is set as a percentage of S∞. If this
lower limit is specified as a number greater than one, it actually becomes an upper limit
specification.
____________________________________________________________________________________________

Page 6 of 18
With above parameters, a wide range of shape specifications becomes possible.

Figure 4 lists a range of potential shapes including some comments as the
circumstances these could be appropriate.
Figure 4: Selection of Term Structure Shapes
250 250 Slope Slope Lower

Slope Slope Lower
S_indef T_indef (t=0) (t=T∞) a4 limit S_indef T_indef (t=0) (t=T∞) a4 limit
200 200 50 1.5 150 0 -20 0
125 0 0 0 0 0
Credit Spread
Credit Spread
150 150
100 100
T_indef
50 50
0 0
0 2 4 6 8 0 2 4 6 8
Years to Maturity Years to Maturity
Chart 1: Simple, constant spread Chart 2: No hump, suitable for high quality bonds
250 Slope Slope Lower 500
S_indef T_indef (t=0) (t=T∞) a4 limit
200 125 2.5 150 0 10 0 400
T_indef
T_indef
Credit Spread
Credit Spread
150 300
100 200 Slope Slope Lower
S_indef T_indef (t=0) (t=T∞) a4 limit
50 100 270 3.5 600 -5 10 0
0 0
0 2 4 6 8 0 2 4 6 8
Years to Maturity Years to Maturity
Chart 3: Small hump, suitable for medium Chart 4: Pronounced hump, suitable for non-
quality bonds investment grade bonds
There is a final question not answered yet? How would the user actually determine
appropriate parameters for the specific market conditions?
An academic approach would entail estimating them based on observed market

spreads yet, unfortunately, in practice this is often not feasible. As an example, the
universe of say BB-minus bonds might consist of just four securities which all happen to
mature at almost the same time. A pragmatic solution consists of setting S∞ to the
approximate credit spread of the day but to leave the remaining shape parameters
unchanged. They would only be reviewed if patterns strongly change. The analyst may
even decide to set the spread S∞ not in line with the current market but with his/her own
subjective expectations regarding its future development. S∞ could then be called a
target spread and it is also labelled as such in the practical implementation of the
model.
1.2.3 Detecting cheap/rich bonds: a simplified numerical example
For illustration, the following tables (Table 1 and 2) and figures (Figure 5) present a
simplified numerical example of cheap/rich analysis. There are three bonds analysed.
Bond I has an AA rating while bonds IIa and IIb are rated BBB. with A. A buy [sell]
signal is generated if the model price based on the target spread is above [below] the
market price.
____________________________________________________________________________________________

Page 7 of 18
Table 1: Data Example Bonds

Bond I IIa IIb
Rating AA BBB BBB
Price ($ per $face value)* 102 99 106
Coupon (paid once annually) 4.00% 4.00% 4.00%
Time to Maturity 2.79 yrs 5.20 yrs 7.79 yrs
Yield to Maturity 3.24% 4.22% 3.12%
Spread to Benchmark 136 bps 208 bps 58 bps
Target Spread (derived from
50 bps 122 bps 120 bps
parameters in Table 2)
Model Yield 2.37% 3.36% 3.73%
Model Price (MP) 104.338 103.016 101.761
MP > Price MP > Price MP < Price
Cheap Bond: Cheap Bond: Rich Bond:
Recommendation
Buy Buy Sell
* Price excluding accrued interest
Figure 5: Generic Yield to Maturity Based Cheap Rich Analysis

Yield example
5.0% bonds
4.22%
4.0% 3.73%
3.24%
Yield
3.0% 3.36%
3.12% Model yield
2.0% 2.37% according to
target spread
1.0%
Benchmark curve
0 2 4 6 8 10 (polynomial
Years to Maturity degree 3)
110 Buy Buy Sell

Prices example
106.0 bonds
Bond Price
105 104.34
103.02
102.0 101.76
100 Model prices
99.0 according to target
spread
95
0 2 4 6 8 10
Years to Maturity
150 Bond IIa&b,

Target Spread
Rating BBB
( in bps)
100
50
- Bond I, Rating
AA
0 2 4 6 8 10
Years to Maturity
____________________________________________________________________________________________

Page 8 of 18
Table 2: Term Structure of Credit Spread Parameters

Rating AA BBB
S∞ (in bps) 50 125
T∞ 1 2.5
Slope (t=0) 200 200
Slope (t=T∞) 0 -1
“Hump parameter” a4 -6 2
Lower limit (as % of S∞) 0.8 0.8
Compared to the main model, the version above is simplified in these aspects.
• The illustrative example does not take bid/ask spreads into consideration when
generating buy/sell signals.
• To limit the number of recommendations the user may specify a sensitivity
parameter to suppress recommendations where the price is very close to the
model price. The sensitivity chosen could, for instance, take transaction charges
into account.
• Similarly, for shorter bonds typical price changes in less liquid markets may lead
to very erratic yield moves that do translate into meaningful buy/sell signals. The
user may thus exclude the analysis for very short-term bonds.
• Finally, the main model provides statistics on the credit spreads observed in the
basket. An example is shown in Figure 6 below.
Figure 6: Generic Yield to Maturity Based Cheap Rich Analysis
Analysis credit spreads Swiss industrial bonds 26/02/1999 0:00
Internal Rating AAA AA A BBB BB B Other Grand Total

Number of bonds 1 9 36 31 10 1 10 98
Average spread to Swiss Govt (bps) 4 63 71 125 252 265 74 108
Min of spread (bps) 4 8 20 51 44 265 23 4
Max of spread (bps) 4 175 189 259 915 265 127 915
Years to maturity/ next call (average) 1.8 5.0 4.2 4.5 3.9 6.3 5.9 4.5
500
450
400
350
300
250
200
150
100
50
0
AAA AA A BBB BB B Other Grand Total
Average 4 63 71 125 252 265 74 108
Average Credit Spread by Internal Rating
____________________________________________________________________________________________

Page 9 of 18
2 JP Morgan discount factor model

Redemption yield based models have their limitations. In line with the underlying
assumptions, coupons need to be reinvested at the computed internal rate of return. To
see this, consider the following two hypothetical five year government bonds A and B.
The two bonds differ in their coupon rate, which is 10% for A and 2% for B. A pure yield
to maturity based analysis would call for identical redemption yields yet there would be
a great deal higher reinvestment risk for bond A. For instance, assuming annual coupon
payments, the coupon of bond A, respectively B in year four would need to be
reinvested for one year at that same five year rate. With yield curves generally upward
sloping, this rate is likely to be lower and so the high-coupon bond A will probably
achieve a lower return.
Having said this, yield to maturity based models should nonetheless not be completely
discounted. They are indeed appropriate for many bond markets and instruments that
have following properties:
• Bond coupons are comparably uniform and low (less than 5%)
Since new issues are usually placed at or around par, this is usually the case
when the interest environment has been comparably stable over the preceding
few years.
• The market is not too liquid with comparably wide bid/ask price spreads.
This means the models need be somewhat less discerning to detect arbitrage in
the markets.
Conversely, for liquid bond markets such as the one for US Treasuries, the
shortcomings of pure yield to maturity analysis are too great. Here the proper approach
is to think about bonds as a package of cash flows, more specifically as packages of
zero-coupon instruments. To find, respectively explain the price of the whole package,
each of this zero bonds is then discounted at a unique interest rate appropriate for the
time period in which the cash flow will be received.
The model shown here is known under the name “JP Morgan (JPM) Discount Factor
Model”. A preliminary literature search revealed that a description of this model could
not be found in standard academic data bases. Having said this, it might have been
published under a different name.
The JPM Discount Factor Model refrains from modelling the unique interest rates at
each point in time because these are, in colloquial terms, “not very well behaved”.
Experience shows that interest term structures come in all kinds of shapes. There are
not just upward or downward sloping curves but they often have “humps” and “kinks”
that make the formulation of an suitable model very demanding. On the other hand,
modelling the discount factors is an easier approach. This function has a clear boundary
at time zero and is monotonously declining over time. In the JPM model it is modelled
as a polynomial with coefficients determined such that the sum of least square errors of
market price minus model price is minimized.
The following first shows the mathematical derivation of the model which is followed by
a brief description of its implementation for the small basket of NZ government bonds.
____________________________________________________________________________________________

Page 10 of 18
2.1 Mathematical derivation
2.1.1 Fitting the coefficients
Bond universe with n bonds with market prices P = [p1,p2, … , pn]T which all must be
equal to the present value of expected cash flows:
c1dt1,1 + c1dt1, 2 + (1 + c1 )dt 1, 3

= p1
c2 dt 2 ,1 + c2 dt 2 , 2 + c2 dt 2 , 3 + (1 + c2 )dt 2,4
= p2
: : : : :
cn dt n ,1 + cn dt n , 2 + ... ... + cn dt n , j + ... ... + (1 + cn )dt n ,n = pn
with
ci : coupon rate i of bond 1…n.
d ti , j : discount factor at the time of the jth coupon if bond i.
dt i , j is set to the polynomial a m t im, j + a m −1t im, j−1 + ... + a1t i , j + a0 .
The method finds the vector of coefficients A = [a0 , a1...am ] that minimizes the sum of
T
the squared difference between market price vector P and model price vector
PM = [ pm0 , pm1... pmm ]
T
The solution of this least square problem is derived for a three dimensional case m=3:
 c1 c1 ∑ t1, j c1 ∑ t12, j c1 ∑ t13, j 

 j j j   pm1 
c c2 ∑ t2, j c2 ∑ t22, j c2 ∑ t23, j  a0   
 2 j j j     pm1 
c a
c3 ∑ t3, j c3 ∑ t32, j c3 ∑ t33, j  ×   =  pm3 
1
 3 a   
 j j j
  2  : 
: : : :   a3 
 pmn 
cn cn ∑ tn , j cn ∑ tn2, j cn ∑ tn3, j 
 j j j 
n
∂ ∑ ( pmi − pi )
2
n
Minimize ∑ ( pm − p )
i =1
i i
2
⇒ i =1
∂ak
= 0 for k = 0,1,2,3 yields four equations for
the four unknown a0 , a1 , a2 , a3 .
____________________________________________________________________________________________

Page 11 of 18
 c1 c1 ∑ t1, j c1 ∑ t12, j c1 ∑ t13, j 

 C1,1 C1, 2 C1,3 C1, 4   j j j 
C C2 , 2 C2 , 3

C2, 4  c2 c2 ∑ t2, j c2 ∑ t22, j c2 ∑ t23, j 
Defining C= 
2 ,1
= j j j ,
 : : : :  : : : : 
   
Cn ,1 Cn , 2 Cn , 3 C n , 4  c
n cn ∑ tn , j cn ∑ tn2, j cn ∑ tn3, j 
 j j j 
… one must solve the system of following linear equations to find the coefficient vector
A:
CT C × A = CT × P ⇒ A = CT C( ) × (C
−1 T
×P )
The model prices are then found by multiplying C with the coefficient vector A:
C × A = PM
Bonds priced below [above] their corresponding model price are “cheap” [“rich”].
The vector of (under pricing)/over pricing R = [r0 , r1 ...rm ] is then:
T
R = PM − P = C × A − P
2.1.2 Boundary conditions
Often constraints are introduced to force the discount factor at time zero to one (which
is a most reasonable assumption), respectively one sets the first derivative at time zero
to fit a short-term rate, e.g. overnight bank rate observed in the market.
For the restriction d (t=0) =1 ao is set to 1.
For an assumed short-term rate r, a1 becomes minus the continuously compounded

equivalent of the short rate, respectively in terms of the annual equivalent rate:
a1= - rc = - ln(1 + rann)
where
rann is the short-term rate expressed as an annual equivalent yield
rc is the short-term rate expressed as an continuously compounded yield
Derivation:
The discount factor at time (t0 + ∆t ) for rc continuously compounded yield with Taylor
expansion of exponential function
1
a0 + a1 (t0 + ∆t ) + a2 (t0 + ∆t ) + ... = e− (t 0 + ∆t )rc = e−t 0 rc − rc e−t 0 rc ∆t + rc2e −t 0 rc ∆t 2 + ...
2
2
for t0 = 0 , a0 = 1 and omitting second and higher order terms of ∆t :
1
1 + a1 ∆t + a 2 ∆t 2 + ... = 1 − rc ∆t + rc ∆t 2 + ... ⇒ a1 = − rc = − ln(1 + rann )
2
____________________________________________________________________________________________

Page 12 of 18
2.2 Description EXCEL/VBA model parameters
The JPM model is saved under the name
Term structure JP Morgan Model (Feb04).xls
It applies the model to the small universe of NZD government bonds. It could easily be
adapted to determine cheap, respectively rich bonds for the universe of any basket of
uniform bonds.
JP Morgan Discount Factor Modelling

Settlement date
14-Feb-99
Coupon Maturity Bid Ask Mid JPM Fair Price (cheap) / rich JPM Coefficients
6.50% 15-Feb-00 100.563 100.583 100.57% 101.17% (0.60%) a0 1
8.00% 15-Feb-01 102.786 102.854 102.82% 104.48% (1.66%) a1 -0.04879016
10.00% 15-Mar-02 108.406 108.526 108.47% 111.34% (2.87%) a2 -0.00222866
5.50% 15-Apr-03 96.673 96.827 96.75% 97.27% (0.52%) a3 0.000197076
8.00% 15-Apr-04 105.034 105.234 105.13% 106.56% (1.43%) a4 #N/A
8.00% 15-Nov-06 106.518 106.809 106.66% 106.06% 0.61% a5 #N/A
7.00% 15-Jul-09 100.549 100.903 100.73% 98.91% 1.81% a6 #N/A
6.00% 15-Nov-11 91.666 92.049 91.86% 92.83% (0.97%) a7 #N/A
Model Parameters
deg 3 Degree JP Morgan polynomial

restr 0.05 0: no restrictions, 1:DF(t=0) =1, other values: short rate
frequency 2 Number of coupon payments per year
Discount Function Zero Rates

8.0%
1 7.0%
0.8 6.0%
5.0%
0.6
4.0%
0.4 3.0%
2.0%
0.2
1.0%
0 0.0%
- 2 4 6 8 10 12 - 2 4 6 8 10 12
Parameters as well prices are set in yellow shaded areas.

The major parameters are ..
deg: This allows the choice of the degree of the discount function polynomial.
It is not meaningful to choose a much larger value for the small bond universe
in the example.
restr: 0 means no restrictions,
1 means discount function is forced to one for t=0 which means coefficient ao is
set to one.
any other value is assumed to be the short-term rate (annual effective yield,
i.e. non-continuous yield). a1 is calculated with the formula given earlier in
section “Boundary conditions”. At the same time, the discount function and thus
ao is also forced to one.
Price data would have to be linked to a real time data provider and need to be specified
without accrued interest. In the New Zealand market bonds are quoted on a yield to
____________________________________________________________________________________________

Page 13 of 18
maturity basis. Such yields would first need to be converted into prices using the RBNZ
formula:
1  n C  FV 
P=  ∑ k 
+ n
  k =0 (1 + i )  (1 + i ) 
a
(1 + i ) b
where
P: Market Value of Bond
FV: Nominal or face value of bond
i: Annual market yield / 2 (in %)
c: Annual coupon rate in %
C: Coupon Payment (= c/2 * FV) – semi-annual coupon
n: number of full coupon periods remaining until maturity
a: Number of days from settlement to next coupon date
b: Number of days from last to next coupon date
____________________________________________________________________________________________

Page 14 of 18
3 References
Black, F. Scholes, M. (1973). "The pricing of options and corporate liabilities". Journal of
Political Economy 81. p. 637-654
Fabozzi, F.J. (1999). “Bond Markets, Analysis and Strategies” 4th Edition. Prentice Hall,
Englewood Cliffs, N.J.
Krippner, L. “The OLP model of the yield curve: a new consistent cross-sectional and
inter-temporal approach”. Conference Paper NZAE June 2002. Retrieved from
http://nzae.org.nz/files/%2349-KRIPPNER.pdf February 2004
Longstaff, F.A. Schwartz, E.F. (1995) "A Simple Approach to Valuing Risky Fixed and
Floating Rate Debt". Journal of Finance, July 1995, pp.789-819.
Merton, R.C. (1974). “On the pricing of corporate debt: The risk structure of interest
rates”. Journal of Finance 29, pp. 449–470.
Press, W.H. Flannery, B. P. Teukolsky, S. A. Vetterling, W.T. (1992). “Numerical recipes

in C: the art of scientific computing”. Cambridge University Press.
Also available online under http://www.nr.com/
Reilly, F. K. Brown, K. C. (2002) ”Investment Analysis and Portfolio Management”.

South-Western College Pub; 7th edition (October 18, 2002)
Vasicek, O. (1977). "An Equilibrium Characterization of the term structure." Journal of

Financial Economics 5: 177-188.
____________________________________________________________________________________________

Page 15 of 18 Appendix 1
Appendix 1: Longstaff & Schwartz 1995 Model
In their 1995 Journal of Finance paper, Francis A. Longstaff and Eduardo S. Schwartz
(L&S 95) show a simple approach to value risky debt subject to both default and interest
rate risk. In line with the traditional Black-Scholes (1973) and Merton (1974) contingent
claims-based framework, default risk is modelled using option pricing theory. This
means default occurs if the level of asset of a firm (V) falls below a bankruptcy threshold
(K). V is assumed the follow the following stochastic process
dV = µVdt + σVdZ1
where σ is the instantaneous standard deviation of the asset process (constant) and
dZ1 is a standard Wiener process.
Similarly, interest rates are assumed to follow a standard Vasicek process (Vasicek
1997) as follows:
dr = (ζ − βr )dt + ηdZ 2
where
ζ is the long-term equilibrium of mean reverting process (constant)
β is the "pull-back" factor - speed of adjustment (constant)
η is the spot rate volatility (constant)
dZ 2 is a standard Wiener process.
The instantaneous correlation between dZ1 and dZ 2 is ρdt
L&S 95 derive the value of a risky discount bond as
P ( X , r , T ) = D (r , T ) − wD (r , T )Q ( X , r , T )
Thus the price of this bond is a function of X, which corresponds to the ratio of V/K, the
interest rate r, and the time to maturity T. The terms to be calculated are explained
below.
D(r , T ) = e A(T )− B (T ) r ) is the value of riskfree (no credit risk) discount bond according to
Vasicek (1977) with
A(T ) = η
 2
(
α T + η 3 − α 2  e − βT − 1 − η ) (
 − 2 βT
)
2 2
2 − 3 e −1
 2 β β   β β   4 β 
1 − e − βT
B (T ) =
α
Here α represents the sum of the parameter ζ plus a constant representing the market
price of risk, β is defined above
The Q(X,r,T) term can be interpreted as probability - under risk neutral measure - that
default occurs. It is the limit of Q( X , r , T , n) as n → ∞ .
Q( X , r , T , n) is calculated as follows:
____________________________________________________________________________________________

n
q1 = N (a1 )
Q( X , r , T , n) = ∑ qi with i −1
i =1 qi = N (ai ) − ∑ q j N (bi j ), i = 2,3,....., n

j =1
where N(.) denotes the cumulative standard normal distribution and
− ln X − M (iT n, T ) M ( jT n, T ) − M (iT n, T )
ai = bi j =
S (iT n ) S (iT n ) − S ( jT n )
Expressions M (t , T ) and S (T ) are defined as follows:

 α − ρση η 2 σ 2 
M (t , T ) =  − 2− t
 β β 2   ρση η 2 
S (T ) =  + 2 + σ 2 t
 ρση η 2   β β 
+  2 +  exp(− βT )(exp(βt ) − 1)
 β 2β 3   ρση 2η 2 
and −  2 + 3 (1 − exp(− β t ))
 r α η2   β β 
+  − 2 + 3 (1 − exp(− βt ))
β β β   η2 
+  3 (1 − exp(− 2 β t ))
 η2   2β 
−  3  exp(− βT )(1 − exp(− βt ))
 2β 
As a reminder, ρ is the instantaneous correlation between the asset and interest rate
processes.
Finally, the constant parameter w is the write-down in case of a default in percent of the
face value. In other words, it is one minus the recovery rate in case of a default.
Once the value of a pure discount bond is found, the value of a coupon bond is simply
valued as series of discount bonds consisting of coupons and principal repayment. Note
that L&S 95 also derive a closed form solution for perpetual floating rate debt in a
similar fashion.
The authors conclude their work with an empirical model validation. They conduct a
regression analysis of how historically observed spreads (sourced from Moody’s) have
correlated with the return of share indices as a proxy for the asset process, respectively
change in interest rates. They indeed find significant negative correlations in most cases
with both interest rate changes and development of asset prices. Just for high grade
bonds (AAA and AA bond) they determined less significance for the asset correlation
coefficient. This may be expected though because well cushioned high grade credits will
be less affected by downturns in the share market.
As an illustration of the L&S 95 model outputs, the charts on the following page show
the value and yield of a discount bond as a function of time to maturity T for the
parameters in Table A1.
____________________________________________________________________________________________

Chart A1: Risky discount bond prices as a function of bond tenor (time to
maturity)
Value risky discount bond

(L&S 95)
100%
Value risk free discount bond
90%
(Vasicek)
80%
70%
60%
50%
40%
30%
20%
10%
0%
0 2 4 6 8 10 12
Time to maturity
Chart A2: Term Structure of Interest Risky Discount Bonds

Term Structure of Interest Risky Discount Bonds
16%
14% Risky Discount

Bond
12%
10%
8%
r at t=0
6%
4% Yield riskfree bond

2%
0%
0 2 4 6 8 10 12
Time to maturity
____________________________________________________________________________________________

Table A1: Parameters L&S 95 Model Example

Parameter description Symbol Value
Rate r0 at t=0 r 7.0%
"Pullback" factor interest rate β 1.00
Instantaneous std dev. of short rate η 3.162%
η2 0.0010
α in L&S = ζ + constant α 0.060
V0/K =X (measure of initial credit quality) X 1.30
Writedown = 1 - Recovery Rate w 0.50
Volatility of asset value process σ 20.00%
σ2 0.0400
Instantaneous correl. asset/interest rate ρ - 0.25
Iterations for Q n 100
____________________________________________________________________________________________


Relative Value Models (Feb04)

Uploaded by

Copyright:

Available Formats

Relative Value Models (Feb04)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Relative Value Models (Feb04)

Uploaded by

Copyright:

Available Formats

What are the two relative value models described in the document?

What are the two relative value models described in the document?

What are the components of the simple yield to maturity based model?

What are the components of the simple yield to maturity based model?

Bond Relative Value Models

The detailed link is currently

Kurt Hess, [email protected] 23/02/2004

1 Simple yield to maturity based benchmarking

• ConfBenchmark (Feb04).xls shows the implementation of a redemption yield

• Cheap Rich List (Feb04).xls produces a cheap/rich analysis for a universe of

1.1 Benchmark curve

Figure 1: Generic Time / Yield Chart with Bond Yield Curve

Kurt Hess, [email protected] 23/02/2004

The following paragraph derives the calculation of the polynomial coefficients.

accuracy. However, this is not an issue in most cases because meaningful

Kurt Hess, [email protected] 23/02/2004

1.2 Redemption yield based cheap rich analysis

An example: a five year government bonds yields 4.5%, a corresponding corporate

1.2.1 The nature of credit spreads

The small chart to the right

for short-term bonds the

Kurt Hess, [email protected] 23/02/2004

1.2.2 Defining the term structure of credit spreads in the model

Figure 3: Explanation Shape Parameters Term Structure of Credit Spreads

100 Spread Lower Limit

50 Short-term characteristics long-term

It must meet the following three boundary conditions:

Kurt Hess, [email protected] 23/02/2004

With above parameters, a wide range of shape specifications becomes possible.

Figure 4: Selection of Term Structure Shapes

250 250 Slope Slope Lower

An academic approach would entail estimating them based on observed market

1.2.3 Detecting cheap/rich bonds: a simplified numerical example

Kurt Hess, [email protected] 23/02/2004

Table 1: Data Example Bonds

Figure 5: Generic Yield to Maturity Based Cheap Rich Analysis

110 Buy Buy Sell

150 Bond IIa&b,

Kurt Hess, [email protected] 23/02/2004

Table 2: Term Structure of Credit Spread Parameters

Figure 6: Generic Yield to Maturity Based Cheap Rich Analysis

Analysis credit spreads Swiss industrial bonds 26/02/1999 0:00

Internal Rating AAA AA A BBB BB B Other Grand Total

Average Credit Spread by Internal Rating

Kurt Hess, [email protected] 23/02/2004

2 JP Morgan discount factor model

Kurt Hess, [email protected] 23/02/2004

2.1 Mathematical derivation

2.1.1 Fitting the coefficients

c1dt1,1 + c1dt1, 2 + (1 + c1 )dt 1, 3

dt i , j is set to the polynomial a m t im, j + a m −1t im, j−1 + ... + a1t i , j + a0 .

 c1 c1 ∑ t1, j c1 ∑ t12, j c1 ∑ t13, j 

the four unknown a0 , a1 , a2 , a3 .

Kurt Hess, [email protected] 23/02/2004

 c1 c1 ∑ t1, j c1 ∑ t12, j c1 ∑ t13, j 

2.1.2 Boundary conditions

For the restriction d (t=0) =1 ao is set to 1.

For an assumed short-term rate r, a1 becomes minus the continuously compounded

a1= - rc = - ln(1 + rann)