Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

A Comparison of Bootstrapped Forward Rates Using Different Discounting Factors

Michael Kinnally January 26, 2012


Let t0 , t1 , ..., tN partition a period of time into quarters, i.e., the ith quarter lies in the interval of time [ti1 , ti ]. Assume we know the values of the quarterly forward rates f1 , ..., fN 1 , where for each i = 1, ..., N 1, fi is the forward rate over the period [ti1 , ti ], and let i be the year fraction over the same period. Assume we know the quoted rate of r for the xed leg of a swap where each leg pays quarterly. This yields the equation
N N

i rDFi =
i=1 i=1

i fi DFi ,

(1)

where DFi is the risk free discount factor up to time ti ; this may be based on a term structure independent from f1 , ..., fN . This swap equation can be solved for fN as follows:
N 1

fN = r +
i=1

ci

DFi , DFN

i (r fi ) for i < N. N If the discount factors are tied to the forward rates F1 , ..., FN , then they can be written as ci :=
i

where

DFi =
j=1

1 , 1 + j Fj

so that

DFi = DFN
N

(1 + j Fj ),
j=i+1

from which it follows that fN = r +

N 1

ci
i=1

(1 + j Fj ).
j=i+1

u u Let F1 , ..., FN and F1 , ..., FN be two term structures of forward rates over the given quarters, and u u let DF 1 , ..., DF N and DF 1 , ..., DF N be the corresponding discount factors, i.e., for each i = 1, ..., N , i

DF u = i
j=1

1 1 + j Fju

and

DF i =

1 . 1 + j Fj j=1

Assume that the rst term structure dominates the second, i.e., for each i, Fiu Fi , 1 (2)

which implies that 1 + i Fiu 1 + i Fi , and thus r > fi ) for each i, then
N 1 u fN := r + i=1

DF u i DF u N

>

DF i DF N

for each i. Therefore, if ci > 0 (i.e.,

ci

DF u i > r+ DF u N

N 1

ci
i=1

DF i =: fN , DF N

u and conversely, if ci < 0 (i.e., r < fi ) for each i, then fN < fN . Unfortunately, equation (1) makes it very unlikely for ci to have constant sign, since i DFi > 0 for all i. For example, if each ci > 0, then this would imply a much larger nal forward rate fN to make up for the consistently higher xed rate r. However, even if ci is not always above or below zero, we can still have an idea about the effects of the different term structure. Assuming (2), we have for i1 < i2 that

DF i2 DF u2 i u DF N DF N

= <
N N

(1 + j Fju )
j=i2 +1 N

(1 + j Fj )
j=i2 +1

(1 + j Fj )

i2

(1 +

j Fju )

i2

(1 + j Fj )
j=i1 +1

j=i2 +1

j=i2 +1 N

=
N

(1 + j Fju )
j=i2 +1

(1 + j Fj )
j=i1 +1 N

(1 + j Fj )
j=i1 +1

<

(1 + j Fju )
j=i1 +1

(1 + j Fj )
j=i1 +1

DF u1 DF i1 i . u DF N DF N

Therefore, the sign of ci (or equivalently, the sign of r fi ) has more of an effect for smaller i. For example, if N = 3, and (r f1 ) = (r f2 ) > 0 and 1 = 2 = 3 = , then
u u f3 = r + (r f1 )(1 + F2 ) + (r f2 ) u = r + (r f1 )F2 u > r + (r f1 )F2

= f3 .

You might also like