Eco 423 Exam 2018 Sol
Eco 423 Exam 2018 Sol
Eco 423 Exam 2018 Sol
with the interpretation that the ‘convenience’ over a period of length dt is given by
δt St dt. The rate δt has a correlation with the rate of return on the market portfolio of
ρ. Equation (2) has the solution
Z t
−kt −kt
δt = δ0 e + δ̄(1 − e ) + v e−k(t−u) dWuδ .
0
Assume the two standard Wiener processes W and W δ are independent. Denote the
Wiener processes under Q by W̃ and W̃ δ respectively.
Problem 1 (50%)
a) What is the economic interpretation of µ, σ, k, δ̄, and v? Moreover, give an
intuitive motivation for what the “price of risk” or “Sharpe ratio” of S is (λSt ; over
a time interval of length dt).
Answer: µ is the growth rate of the commodity. The total compensation—
expected rate of return—is the sum of price/“capital” gains and “dividend” pay-
outs, µ + δt , and is determined by the systematic risk of the commodity. σ and v
capture the total risk of the commodity price and its convenience yield. δ̄ is the
long-run mean of δ, and k determines how quickly the process reverts towards δ̄.
The Sharpe ratio of S is thus given by (µ + δt − r)/σ, as the expected rate of return
is given by µ + δt .
b) Compute E {δt } and Var (δt ), and be clear about what properties/results you use
in your intermediary calculations.
2
Answer: To compute the expectation, it is useful to view the stochastic integral
as a discrete sum
Z t nX o
−k(t−u) δ −k(t−ui ) δ
E e dWu ≈ E e ∆Wui
0
X
e−k(t−ui ) E ∆Wuδi = 0,
=
where the last equality follows from ∆Wuδi ∼ N(0, ui+1 − ui ). We thus have
For the variance we ignore the deterministic term, and again utilize the discrete
sum
Z t
−k(t−u) δ
Var (δt ) = Var v e dWu
0
X
≈ Var v e−k(t−ui ) ∆Wuδi
X 2
= v2 e−k(t−ui ) Var ∆Wuδi
X Z t
−2k(t−ui )
=v 2
e ∆ui ≈ v 2
e−2k(t−u) du
0
2
t 2
v −2k(t−u) v −2kt
= e = 1 − e .
2k
0 2k
The third transition uses the independence of non-overlapping increments of the
Wiener process, and the fourth transition uses that said increments have a variance
equal to the associated time increment.
c) Describe the SDE for S under the equivalent martingale measure Q when v = 0
(you don’t need to derive it, if you see the answer—but do explain briefly what
you base your claim/insight on).
Answer: When v = 0 the convenience yield becomes deterministic. The system
(1)–(2) is thus equivalent to the Black-Scholes-Merton stock price model with de-
terministic dividend yield that we studied in class. We thus know from derivations
in class that
dSt = (r − δt )St dt + σSt dWt .
d) Derive the SDEs for S and δ under the equivalent martingale measure Q when
v 6= 0. Show in particular that
where SR m is the (constant) Sharpe ratio of the market portfolio, and be clear
about the unstated assumption that causes ρvSR m to appear in (3).
3
Answer: We know that there are no arbitrage opportunities R t iff all discounted
∗ ∗
gains processes are Q-martingales. Let St = St /At and Dt = 0 δu Su /Au du. The
discounted gains process for the commodity is then G∗ = S ∗ + D∗ . From Itô’s
lemma (first transition) and Girsanov’s Theorem (second transition)
Diffusions are martingales iff their drift term equals zero, which in our case requires
µ + δt − r − σλt = 0
iff λ = (µ + δt − r)/σ. Making the substitution dWt = dW̃t − λt dt in (1) yields the
first part of the answer:
Because the convenience yield is risky, we must also risk-adjust (2). Because δ is
not a traded asset we must first estimate the price of risk, which in our case can
be done via the CAPM:
ρvσm
µδ = r + 2 (µm − r)
σm
iff
µδ − r µm − r
λδ = =ρ = ρSR m .
v σm
Thus, by another application of Girsanov’s Theorem
for δ̂ = δ̄ − ρvSR
k
m
, which is a constant. By simple replacement of δ̄ with δ̂ in the
solution to the SDE for δ under P, we get
Z t
−kt −kt
δt = δ0 e + δ̂(1 − e ) + v e−k(t−u) dW̃uδ .
0
4
f ) Let mT = EQ
0 {δT } What is the date 0 present value of the contract that promises
the payment (δT − mT )2 at date T ?
Answer: Given no arbitrage the discounted gains process for this contract must
also be a Q-martingale:
p0 Q pT
=E ,
A0 AT
where pT = (δT − mT )2 , iff
p0 = e−rT EQ (δT − mT )2 .
Notice that the preceding expectation is the expression for the variance of δT under
Q. From our answers to b) and e) we thus have that
v2
p0 = e−rT 1 − e−2kt .
2k
Problem 2 (50%)
Assume for this problem that δ ≡ 0 (i.e., you can ignore it).
a) Write down the Euler scheme for (1) under P and Q (you were asked to write
down the Q dynamics of S in Problem 1 c).
Answer: We have
√
Sn+1 = Sn + an Sn ∆t + σSn ∆tn+1
b) Briefly list the strengths and weaknesses of Value at Risk (VaR), Expected Short-
fall (ES), and the Greeks when used to measure and manage financial risks.
Answer:
– VaR strengths: measures downside risk, rather than down- and upside risk;
easy to compute; easy to understand and communicate; directly related to
“default” probability of banks when tied to capital requirements.
– VaR weaknesses: it’s not subadditive—diversification often yields higher VaR;
it gives incentives for more extreme risk-taking, as it ignores the magnitude
of outcomes in the tail beyond the VaR—it’s a frequency measure.
– ES strengths: measures downside risk, rather than down- and upside risk; easy
to compute; reflects the values of the losses in the tail, rather than just their
frequency; it’s subadditive; it doesn’t cause extreme risk taking incentives.
5
– ES weaknesses: it’s harder to communicate and interpret than VaR
– Greeks strengths: easy to interpret; allows construction of hedging portfo-
lios, that make the total portfolio less sensitive to changes in relevant vari-
ables/parameters.
– Greeks weaknesses: they are ceteris paribus ∼ measures value responses to
one-dimensional changes in the economic environment; requires frequent up-
dating to be efficient; typically doesn’t work when there are large changes in
economic environment ∼ crises; can be computationally costly.
Consider the date 0.5 and date 1.0 standard normal shocks/random variates
where each date 0.5 shock can be followed by either of the date 1.0 shocks—i.e., two
possible states at date 0.5, and four possible states at date 1.0.
Assume that you at date 0 sell one European call option with strike price X = 10 and
time to expiration 1.0, and that the remaining parameter values are S0 = 10, µ = 0.2,
σ = 0.5, and r = 0.05.
c) Do the necessary computations for a VaR analysis at date 0 of the risk of the
position at date 0.5.
Hint: You need to determine the appropriate distribution of call prices at date
0.5.
3.71 5.88 4.26 0.74 0.03 5.80 11.74 0.12 2.02 8.64
3.08 6.98 4.33 1.43 1.09 4.30 0.08 0.12 3.05 11.03
d) Use your assistant’s simulation to determine the 10% VaR and ES for the short
position in the call (using weak inequalities in the defintions of VaR and ES).
Answer: See spreadsheet.
e) Call prices 5.80 and 5.88 correspond to underlying asset prices 15.35 and 15.44.
Use these prices to approximate the call’s delta, ∆, and to construct a ∆-neutral
portfolio. Briefly explain two ways in which you could improve your delta esti-
mate, given the available information. Can both approaches be used for a generic
derivative, not necessarily a European call, or for dynamics different from (1)?
Answer: See spreadsheet for computation of delta.
6
A portolio becomes delta-neutral when its delta is equal to zero. If we “extend”
our short position in the call with n units of the underlying asset, the portfolio
delta becomes
∂
(−c(S) + nS) = −∆ + n.
∂S
Setting this expression to zero and solving for n yields n = ∆. One should thus
buy the number of units reported in the spreadsheet.
One can compute delta for all adjacent prices, and take the average. As the sample
size increases, the standard error will go down, yielding a higher precision estimate.
We can improve this further, since we know that the BSM formula holds under
the current assumptions: simply feed the prices associated with the simulated call
options prices into ∆ = N(d1 ) where d1 = d1 (S). The latter approach does not
work in general though, when closed form formulas are not available.