Scribd
Upload
16 views

Assignment One

Uploaded by

henrychenace
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
16 views

Assignment One

Uploaded by

henrychenace
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Finance 3

Assignment Three
Due Date: October 7, 2024.

1 Theoretical Problems
1. Let Xt be a diffusion process
dXt = µ(Xt )dt + σ(Xt )dBt
where B is a standard Brownian motion.

(a) Let f a bounded twice continuously differentiable function with


bounded derivatives. Let A be the generator of the process Xt that
you identified in the previous question. Verify that each of the
following processes are martingales with respect to the filtration
generated by Xt .
Rt
i. Mt = f (Xt ) − 0 (Af )(Xs ) ds
Rt
ii. Nλ,t = e−λt f (Xt )+ 0 e−λs (λf (Xs )−Af (Xs )) ds, where λ ∈ R.
 R 
t
iii. If inf x f (x) > 0: Rt = f (Xt ) exp − 0 Af (Xs )
f (Xs )
ds
(b) Let two numbers c and d be such that c < x < d, where x = X0 .
By choosing a suitable function f show that:
Rd  R 
z 2µ(y)
x
exp − 2
c σ (y)
dy dz
P(Xt reaches c before d) = R d  R
z

c
exp − c 2µ(y)
σ 2 (y)
dy dz

2. Let Xt be a diffusion process


dXt = µ(Xt )dt + σ(Xt )dBt
where B is a standard Brownian motion, and σ > 0 and µ are smooth
bounded functions. Let c < d, let τ = inf{t > 0|Xt ∈
/ (c, d)}. Suppose
that f solves the differential equation:
σ 2 (x) ′′
f (x) + µ(x)f ′ (x) = −1, f (c) = f (d) = 0. (1)
2
Show that Ex [τ ] = f (x).

1
3. Consider the following boundary value problem in the domain [0, T ]×R.

∂F ∂F σ 2 (t, x) ∂ 2 F
+ µ(t, x) + + k(t, x) = 0
∂t ∂x 2 ∂x2
F (T, x) = G(x)

Here µ, σ, k and G are assumed to be known functions. Prove that this


problem has the stochastic representation formula:
Z T
F (t, x) = Et,x [G(XT )] + Et,x [k(s, Xs )] ds
t

where as usual X has the dynamics

dXs = µ(s, Xs ) dt + σ(s, Xs ) dWs


Xt = x.

4. Solve the problem of maximizing logarithmic utility


Z T 
−δt
E e log(ct ) dt + K · log(XT )
0

given the usual wealth dynamics

dXt = Xt u0t r + u1t α dt − ct dt + u1t σXt dWt ,


 

and the usual control constraints

ct ≥ 0, ∀t ≥ 0,
u0t + u1t = 1, ∀t ≥ 0. (2)

5. A numeraire for a measure P̃ (equivalent to the initial measure in the


model for a universe of assets) is an asset whose value process satisfies
Nt > 0 for all t, and for which At /Nt is a martingale under P̃ for all
tradable assets At . Show that an investment in the numeraire portfolio
maximizes expected (under P̃ ) log wealth at a given final time T among
all assets with a given current price.

2
6. Consider a general control problem of minimizing
Z T 
u u
E F (t, Xt , ut ) dt + Φ(XT ) (3)
0

given the dynamics

dXt = µ(t, Xt , ut )dt + σ(t, Xt , ut )dWt , (4)

and the constraints


u(t, x) ∈ U. (5)
Now, for any control law u, define the total cost process C(t; u) by
Z t Z T 
u u u
C(t; u) = F (s, Xs , us ) ds + Et,Xtu F (s, Xs , us ) ds + Φ(XT ) ,
0 t
(6)
i.e. Z t
C(t; u) = F (s, Xsu , us ) ds + J (t, Xtu , u). (7)
0
Use the HJB equation in order to prove the following claims.

(a) If u is an arbitrary control law, then C is a submartingale.


(b) If u is optimal, then C is a martingale.

2 Computational Problems
1. Obtain the time series of returns of a stock, returns of a market index
(e.g. the S&P 500 and an interest rate that may be interpreted as
risk-free (e.g. a T-bill rate). Consider the equation:

RSt − rf t = α + β(RM t − rf t ) + εt

where RSt is the return of the stock at time t, rf t is the risk-free rate
at time t and RM t is the return of the market index at time t. Find the
parameters that give the best fit to the above equation by minimizing:

(a) ∥ε∥1 = Tt=1 |εt |


P
qP
T 2
(b) ∥ε∥2 = t=1 εt

3
(c) ∥ε∥∞ = maxt |εt |

2. This exercise asks you to implement a standard backtest that would be


used to analyze a portfolio strategy in practice, for the particular case
of the mean variance model. Write a compute program that accepts
as input the time series of financial returns of a universe of ‘tradeable
instruments’, a window size representing a number of historical obser-
vations to be used for parameter estimation, a set of ‘risk aversions’
τi , i = 1, . . . , M , and the risk-free interest rate. The program should
then start at the beginning of the time series (actually at the nth obser-
vation, if the window size is n observations) and perform the following
steps.

• Estimate the mean vector and variance-covariance matrix of the


returns based on the previous n observations (where n is the win-
dow size).
• Find the mean-variance optimal portfolio wi corresponding to each
τi using the formulation of the mean-variance model with the risk-
aversion parameter in the objective function:

max wT µ − τ wT Σw (8)
w
wT 1 = 1 (9)

• For the next period, calculate the realized return for each of the
portfolios wi that were obtained. Also calculate the realized return
of a portfolio that simply places equal weights on each of the
assets.

After completing the above steps, you will have generated a time series
of returns for each investment strategy (one time series corresponding to
each risk aversion level, and one corresponding to the equally weighted
portfolio). Your program should then calculate the following statistics
for each of the strategies based on the empirical distribution (i.e. equal
weight for each point in your return series): expected return, standard
deviation, and Sharpe Ratio.

You may wish to test your program with real financial data, for ex-
ample by downloading time series from yahoo.com, the MSCI indexes

4
available from www.mscibarra.com, or the hedge fund indexes available
from www.hedgefundresearch.com. While it is not required for the ex-
ercise, if you are interested you may wish to investigate other aspects
of the trading strategies (for example, the amount of turnover in the
portfolio from period to period, which is very important in practice
when trades are subject to transactions costs).

You might also like