Scribd
Upload
14 views11 pages

Irm 23

Uploaded by

u
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
14 views11 pages

Irm 23

Uploaded by

u
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

Chapter 2: Estimating the Term Structure

2.3 Smoothing Methods

Interest Rate Models


Damir Filipović
2.3 Smoothing Methods

Exact methods:
• Discount curve is sensitive to small
changes in cash flow matrix.
• Discount factors of similar maturity
can be very different.
• Leads to ragged forward curves.

Smoothing methods: estimate a smooth


forward curve from market rates at the
cost of not exactly matching data.

Interest Rate Models


Abstract Formulation

Set spot date t0 = 0 for simplicity.

Data: N yields yi = y (0, Ti ) with maturities 0 < T1 < · · · < TN .

Aim: find smooth forward curve f (T ) = f (0, T ) matching the yields optimally
Z Ti
f (u) du = Ti yi + i
0

for pricing errors i subject to being minimized.

Interest Rate Models


Key Criteria for Smoothing Methods

Smoothing methods mainly used by central banks.

• Smoothness: supply a market expectation for monetary policy purposes rather


than precise pricing of all bonds in the market.

• Flexibility: sufficiently flexible to capture movements in the underlying term


structure.

• Stability: small changes in data at one maturity do not have disproportionate


effect on forward rates at other maturities.

Interest Rate Models


Estimation Methods Used by Several Central Banks

Bank for International Settlements 2005: Nelson–Siegel (NS), Svensson (S),


weighted prices (wp).

Central bank Method Minimized error


Belgium S or NS wp
Canada Exponential spline wp
Finland NS wp
France S or NS wp
Germany S yields
Italy NS wp
Japan Smoothing spline prices
Norway S yields
Spain S wp
Sweden Smoothing spline or S yields
Switzerland S yields
UK Smoothing spline yields
USA Smoothing spline bills: wp, bonds: prices

Interest Rate Models


Nelson-Siegel and Svensson Curves

Parametric families of forward curves.

• Nelson–Siegel:
f (T ) = β0 + β1 e −aT + β2 aT e −aT
for parameters β0 , β1 , β2 , and a.

• Svensson:

f (T ) = β0 + β1 e −a1 T + β2 a1 T e −a1 T + β3 a2 T e −a2 T

for parameters β0 , β1 , β2 , β3 , and a1 , a2 .

Interest Rate Models


Nelson-Siegel Yield Curves

Nelson–Siegel yield curves are Figure with a = 0.3:

1 T
Z 1.2
y (T ) = f (u) du
T 0 1

= β0 I0 (T ) + β1 I1 (T ) + β2 I2 (T ) 0.8
I 0 (T)

0.6 I 1 (T)
I 2 (T)
with basis functions 0.4

• I0 (T ) = 1 (level) 0.2

1−e −aT
• I1 (T ) = aT
(slope) 0
0 5 10 15 20 25 30
Time to maturity T
1−e −aT
• I2 (T ) = aT
− e −aT (curvature).

Issues: flexibility, stability


Interest Rate Models
Smoothing Splines: Hilbert Space Approach

Find forward curve in Hilbert space H consisting of absolutely continuous


functions on [0, T∗ ] with scalar product
Z T∗
hg , hiH = g (0)h(0) + g 0 (u)h0 (u) du.
0

Solve optimization problem


Z T∗ N 
X Z Ti 2
0 2
min (f (u)) du + α Ti yi − f (u) du (P)
f ∈H 0 0
i=1

where α > 0 tunes trade-off between smoothness and correctness of fit.


Interest Rate Models
Lorimier's Theorem

Quadratic basis splines h1 and h2 together


The unique solution f of (P) is a with first derivatives for T1 = 1 and T2 = 2:
quadratic spline characterized by
N
X 4

h1 (u)
f (u) = β0 + βi hi (u) 3 h1 '(u)
i=1 h2 (u)

2 h2 '(u)
1
where hi ∈ C [0, T∗ ] is a quadratic basis
1
spline with
0

hi0 (u) = (Ti − u)+ ,


0 0.5 1 1.5 2 2.5 3
hi (0) = Ti u

and β0 , . . . , βN solve a linear system.

Interest Rate Models


Lorimier's Theorem Continued

Lorimier’s theorem continued: β0 , . . . , βN solve the linear system of equations


N
X
βi Ti = 0,
i=1
N
!
X
α yi Ti − β0 Ti − βj hhi , hj iH = βi , i = 1, . . . , N.
j=1

Interest Rate Models


Choice of Parameter α

Example: Swiss government bond yields in


Smoothing splines satisfy key criteria: August 2011
smoothness, flexibility, stability.
2

Parameter α tunes trade-off between


1.5
smoothness and correctness of fit:

Yield (%)
Data
• α → 0: maximal smoothness, 1
α = 0.01
α = 0.1
constant forward curve f (T ) = β0 α = 10
0.5

• α → ∞: Rperfect fit subject to


T
minimal 0 ∗ (f 0 (u))2 du 0
0 5 10 15 20 25 30
Time to maturity T

Choice of α is critical.

Interest Rate Models

You might also like