Você está na página 1de 4

DBAnalytics Screening Examination

September 10, 2007

You are being considered for a position in the DBAnalytics group. Please
consider the following screening examination. You must answer one question
from each of the following areas plus two additional questions which you can
choose from the list. The answers should be sent by Email in *.doc or *.pdf
format with your name and date at the top of the document. None of the
questions are trick questions and should yield to standard method of analysis.

1 Rates
1.1 Question 1
Let’s consider a model described by:
df (t, T ) = µf (t, T )dt + σ f (t, T )dWt ,
where dWt is a one dimensional brownian motion, and f (t, T ) is the instant
forward rate.

1. For the model to be arbitrage free, what should µf (t, T ) be equal to ?


2. Show that if the volatility has a separable form (ie σ f (t, T ) can be written
f
σ (t, T ) = g(t)h(T )), diffusing the whole yield curve under that framework is
equivalent to diffusing only one state variable. How could that state variable be
interprated?

1.2 Question 2
Show that a security constitued by a Libor paid in advance has a model inde-
pendant price, that price being a linear combination of caplets (with different
strikes) on the same libor.

2 FX, Equities and Commodities


2.1 Question 1
In an equity market where the realised volatility of the underlying is 25%, some-
one proposes to sell you a call at an implied volatility of 20%.

1
1. Assuming that the realised volatility is going to remain constant at 25%
until the expiry of the option, explain how it is possible to make money from
that opportunity.
2. Let’s now assume that only the expectation of the realised volatility is
25%, and that the number of fixings is big enough. Is it always possible to make
money in that case?

2.2 Question 2
Let F Xt be the EU R/U SD FX rate (U SD is the domestic currency, so that
F X0 is around 1.3), and St be an USD asset.
These two assets have the following dynamics:
dF Xt
 (1) (1)
F Xt = rd − rf dt + σt dWt
St = exp(xt )
(2) (2)
dxt = (φt − λxt ) dt + σt dWt
(1) (2)
< dWt , dWt > = ρdt
(1) (2)
Where rd , rf , λ, ρ are constant, φt , σt and σt are time-dependant and de-
terministic functions. φt is chosen so that the forward curve of St is flat.

1. What should φt be equal to?


2.We consider a call of Strike K and maturity T on S. However the value
+
payoff (ST − K) will be paid in EUR. What is the price of that option ?

3 Credit
3.1 Question 1
Let’s consider a pool of n credits.All the credits have the same notional N and
the same recovery rate R.
Let’s write N (x) the cumulative distribution of a standard normal random
variable N (0, 1)
We assume that there are n credits whose default times τ1 , ..., τn are driven
by normal state variables Y1, ..., Yn as follows:

τi = Fi−1 (N
p(Yi ))
Yi = βZ + 1 − β 2 Xi

were Z, X1 , ..., Xn are independent standard normal random variables, and


Fi is the distribution function of τi with inverse Fi−1 . β ∈] − 1, 1[.

Fi (t) = P (τi < t) = 1 − exp (−λi t) .

B (t, T ) is the value in t of having 1 in T (discount factor).

2
We define a N th to default is a swap by:
- receives N (1 − R) when the Nth credit default
- Pays CP at times t1 , ..., tp if the N th default hasn’t occurred.
(accrued coupons are ignored in the present exercise).

1. Write the price of a First to Default.


2. What is the sensitivity to a change in default intensity λi in one credit ?
3. How would you replicate the receiving leg of a N th to Default with only
First to defaults ?
(all First to default on all subpools are available)

3.2 Question 2
Let’s consider a discrete state credit model.
More specifically that model has N possible states with corresponding mul-
tipliers M1 , ..., MN and probabilities π1 , ..., πN .
The states are static. (no transition between states).
We suppose
N
X
E(M ) = πj Mj = 1.
j=1

Let’s note for the credit:


Z t
Λ (t) = λs ds
0

where λs is the default intensity for the credit.


The probability of the credit defaulting before the time t is:

P (t) = 1 − exp (−Λ (t))

For the credit in the j th state:


Z t
Λj (t) = A (t) + Mj λs ds.
0

Where A(t) is calibrated to get the proper default probability for each time
t:
N
X
P (t) = πj Pj (t)
j=0

where Pj (t) is the probability of the credit defaulting before the time t in
state j.
Show that for all j the default probability in state j increases with time.

3
4 Probability
4.1 Question 1
Let X and Y be two gaussian random variables N (0, σx ) and N (0, σy ). X and
Y are correlated with a correlation ρ.
What is the law of E (X − Y |2X + Y )

4.2 Question 2
Let Bt be a brownian motion. We define the random variable M as M =
max {Bt , t ∈ [0, 1]}, and the stopping time τ = inf {t > 0, Bt >= 1}.
Show that M and √1τ have the same law.

5 C and C++ Programming


5.1 Question 1
Write an int strtoint(char *) program that takes numbers written in a char *
and returns the corresponding int. For example, char * c = ”467” would be
changed to int i = 467.

5.2 Question 2
Let’s consider the following code:
class A
{
...
public:
int compute int() const;
double compute vec( const vector < double > & vec );
};
What is the meaning of ’const’ in these two lines?

5.3 Question 3
A smart pointer is an object that is used as a proxy for a pointer. However it
has the additional feature that it counts how many times the object it points
to is referenced, and destroys it once it’s no longer referenced. Write the code
of a smart pointer.

Você também pode gostar