You are on page 1of 3

Exercise Session 2. Solutions.

Fixed Income and Credit Risk

Question 1. We consider the Leman Bank based in Lausanne. This bank is the world leader of lending to startups.
If a default of a startup occurs, it is assumed that the loss to the bank is distributed as a N µ, σ 2 with µ = 500 000


and σ = 300 000. Defaults are taking place following a Poisson distribution with default intensity of λ = 15 per
year.

• Determine the distribution of a total annual defaulted amount D;



The number of defaults Nt follows a Poisson distrbution. Each loss in default Y is an independent N µ, σ 2 .
Therefore, a total defaulted amount Dt follows a Compound Poisson distribution with a c.d.f.


" k #
X X (λt)k −λt
P [Dt 6 x] = P Yτl 6 x e .
k!
k=1 l=1

The density of a Compound Poisson is obtained by explicit differentiation of the c.d.f. with respect to x,
which yields
∞ 
(λt)k −λt

X 1 − 12
(x−kµ)2
fD (x) = √ e kσ 2 · e .
2πkσ 2 k!
k=1

• Plot the density of D using your favorite software;


For computational details, see file comppois.R in the solution folder. The resulting density plot is below.

Density of a yearly loss

0.0015

0.0010

0.0005

0.0000

0 500 1000 1500 2000

Thousand CHF

• Determine the 5% VaR;


Again, see file comppois.R in the solution folder. You will find the resulting VaR5% on a density plot below.

1
Density of a yearly loss

0.0015

0.0010

0.0005

0.0000 5% VaR: 401.4

0 500 1000 1500 2000

Thousand CHF

• Determine the expected annual loss and its associated standard deviation.
hP i
Nt
For the expected loss, we need to compute E [Dt ] = E i=1 Yτ i , where both Nt and Yτi are random variables.
To deal with each of them separately, use the law of iterated expectations (a.k.a. ‘you can always condition
on more’): "N # " !# "N #
Xt Nt
X Xt

E Yτi = E E Yτi |Nt = k =E E (Yτi ) = µE [Nt ] = µλt.


i=1 i=1 i=1

Above, the one before the last equality follows from an independence of Yτi , and the last step utilizes the
unconditional expectation of a Poisson process that you computed in class with Michael. For given µ = 500 000
and λ = 15, E [Dt ] = 7500 000.
To compute the variance, first express it this way:
"N # 
Nt
!2  "N #!2 
Nt
!2 
Xt X X t X
V Yτi = E  Yτi − E Yτi = E Yτi  − (µλt)2 .
i=1 i=1 i=1 i=1

Then, to compute the expectation of a squared sum of losses, simplify it a bit first:

Nt
!2  
Nt X Nt

X X X
E Yτi  = E Yτi Yτj + Yτ2i  .
i=1 i=1 j6=i i=1

Note that in the expectation above, there are Nt2 − Nt cross-product terms (and the multiplied random
variables are independent from each other in each cross-product) and Nt squared terms. Now, just use the

2
law of iterated expectations again:
     " !#
XNt X Nt
X Nt X
X Nt
X
2 2
E Yτi Yτj + Yτi  = E E  Yτi Yτj |Nt = k  + E E Yτi |Nt = k =
i=1 j6=i i=1 i=1 j6=i i=1
  "N #
XNt X Xt
2
 
= E E Yτi Yτj  + E E Yτi =
i=1 j6=i i=1

= E (Nt2 − Nt )µ2 + E Nt µ2 + σ 2 =
   

= µ2 E Nt2 + σ 2 E [Nt ] =
 

= µ2 λt + (λt)2 + σ 2 λt.


Combining with the previous result, we find:


"N #
Xt

Yτi = λt µ2 + σ 2 .

V
i=1

For given values of µ = 500 000, σ = 300 000, λ = 15, V [Dt ] = 51 · 109 .

Question 2: Demonstrate that


Z b N
X
xdx = lim xi−1 (xi − xi−1 ),
a N →∞
i=1

b−a
assuming (to make your life a bit easier) that xi − xi−1 = N for all i, i.e., the xi are equally spaced.
To start with, you can compute the left-hand side explicitly to find
Z b
1 2
b − a2 ,

xdx =
a 2

and now you need to compute the right-hand side to arrive to the same expression.

N N  
X X xi + xi−1 xi − xi−1
lim xi−1 (xi − xi−1 ) = lim − (xi − xi−1 ) =
N →∞
i=1
N →∞
i=1
2 2
N N
X x2i − x2i−1 1X
= − (xi − xi−1 )2 =
i=1
2 2 i=1
1 2  1 (b − a)2
= b − a2 − lim =
2 2 N →∞ N
1 2
b − a2 .

=
2

You might also like