Professional Documents
Culture Documents
Lecture 7 - 2023 (1-55, 66-68)
Lecture 7 - 2023 (1-55, 66-68)
Mark-Jan Boes
March 1, 2023
Part I
Introduction
Lecture 7
Mark-Jan Boes 1
Motivation
Lecture 7
Mark-Jan Boes 2
Motivation
If the risk turns out to unacceptable, then there are two choices:
1 lower the ambition
2 change the attitude towards risk
Lecture 7
Mark-Jan Boes 3
Motivation
Basic principles:
risks that do not offer a compensation should be avoided (i.e.
hedged) as much as possible
all risks of investing in a financial instrument should be well
understood
the risk premium of a financial instrument should be in line
with business / personal preferences
only invest in well-diversified portfolios, i.e. avoid unnecessary
specific risks
Ok, makes sense, but what are the tools that we can use to decide
on the investment of EUR 4 bln?
Lecture 7
Mark-Jan Boes 4
Return distribution volatility product
35.0%
30.0%
25.0%
probability
20.0%
15.0%
10.0%
5.0%
0.0%
<-15%
-15% - 12.5%
-12.5% - -10%
-10% - -7.5%
-7.5% - -5%
-5% - -2.5%
-2.5% - 0%
0% - 2.5%
2.5% - 5%
5% - 7.5%
7.5% - 10%
10% - 12.5%
>15%
monthly return (USD)
Lecture 7
Mark-Jan Boes 5
Portfolio Choice
Lecture 7
Mark-Jan Boes 6
Coming lectures
Lecture 7
Mark-Jan Boes 8
Academic approach
Lecture 7
Mark-Jan Boes 9
Academic approach
Lecture 7
Mark-Jan Boes 12
Academic approach
The conditional mean and conditional variance are the mean and
variance conditional on the investor’s information at time t, thus
they are written with t subscripts.
Lecture 7
Mark-Jan Boes 13
Academic approach
The investor puts a share wt of his portfolio into the risky asset.
Portfolio characteristics:
P
Et (Rt:t+1 f
) = Rt:t+1 f
+ wt (Et (Rt:t+1 ) − Rt:t+1 )
2 = w 2σ2
σpt t t
Lecture 7
Mark-Jan Boes 14
Academic approach
Lecture 7
Mark-Jan Boes 15
Academic approach
Σ−1
t µt
wt∗ =
γ
where:
Σt is the variance-covariance matrix of the returns on risky
assets
µt is the time-t vector of expected excess returns on risky
assets
γ is risk aversion
Obvious consequence: if γ = ∞ then the weight in risky assets is
zero and hence all money is invested in the risk free asset.
Lecture 7
Mark-Jan Boes 16
Mean variance approach
Let us first stay in the 1-period setting where the single period
covers a small time interval.
Lecture 7
Mark-Jan Boes 17
Power utility
Lecture 7
Mark-Jan Boes 18
Power utility
1
E(U(W (T ))) ≈ U(E(WT )) + 0 + U 00 (E(WT ))Var (WT )
2
1 000
+ U (E(WT ))E(WT − E(WT ))3 + ...
6
Given that the power utility function is concave, the formula shows
that the variance of wealth contributes negatively to expected
utility.
The formula also shows that higher order moments of the wealth
distribution at time T are important for expected utility and
therefore for optimal portfolio choice.
unless of course the wealth distribution is normal
Lecture 7
Mark-Jan Boes 19
Power utility
Lecture 7
Mark-Jan Boes 20
Power utility
1−γ 1
max log(Et (Wt+1 )) = (1 − γ)Et (wt+1 ) + (1 − γ)2 σwt
2
αt 2
The budget constraint can be written in log form as:
P
wt+1 = rt:t+1 + wt ,
P
where rt:t+1 P
= log(1 + Rt:t+1 ), the continuously compounded
portfolio return.
P 1 2
max Et (rt:t+1 ) + (1 − γ)σpt
αt 2
Lecture 7
Mark-Jan Boes 21
Power utility
Lecture 7
Mark-Jan Boes 22
Power utility
But how can we decide on the optimal portfolio for a long term
investor?
The long term investor’s optimal portfolio depends not only on his
objective but also on what he is allowed to do each period.
Lecture 7
Mark-Jan Boes 25
Long term portfolio choice: the buy-and-hold investor
What can we say about the optimal portfolio for the buy-and-hold
investor?
Lecture 7
Mark-Jan Boes 26
Long term portfolio choice: the buy-and-hold investor
All these assumptions imply that we can adjust the formula derived
earlier for K periods (slide 23):
f
Et (rt:t+K ) − rt:t+K + 12 σKt
2
wt∗ = 2
γσKt
f
K Et (rt:t+1 ) − Krt:t+1 + 21 K σt2
=
K γσt2
f
Et (rt:t+1 ) − rt:t+1 + 12 σt2
=
γσt2
Lecture 7
Mark-Jan Boes 27
Asset allocation in an asset-only framework
Lecture 7
Mark-Jan Boes 28
Barberis: analysis 1
Observations:
optimal allocation in equity depends heavily on risk aversion A
optimal allocation in equity does not depend on horizon
Lecture 7
Mark-Jan Boes 29
Barberis: analysis 1
Lecture 7
Mark-Jan Boes 30
Barberis: analysis 1
Conclusion:
Horizon does not matter for optimal investing under the following
conditions:
power utility
returns are independent and identically distributed
risk attitude does not vary with horizon
full knowledge on the model parameters
Lecture 7
Mark-Jan Boes 31
Barberis: analysis 1
Lecture 7
Mark-Jan Boes 32
Barberis: analysis 1
rt:t+h = µ + εt:t+h
where
εt:t+h ∼ N(0, σ 2 )
Lecture 7
Mark-Jan Boes 33
Barberis: analysis 1
Lecture 7
Mark-Jan Boes 34
Barberis: analysis 1
Consequently:
Lecture 7
Mark-Jan Boes 35
Barberis: analysis 1
This interval is huge: statistically, you cannot reject that the true
µ differs from 0.15%.
Lecture 7
Mark-Jan Boes 36
Barberis: analysis 1
Lecture 7
Mark-Jan Boes 37
Barberis: analysis 1
Lecture 7
Mark-Jan Boes 38
Barberis: analysis 1
Lecture 7
Mark-Jan Boes 39
Barberis: analysis 1
Results:
Lecture 7
Mark-Jan Boes 40
Barberis: analysis 1
Lecture 7
Mark-Jan Boes 41
Asset allocation in an asset-only framework
Lecture 7
Mark-Jan Boes 42
Barberis: analysis 2
zt = a + Bxt−1 + t
where
zt = (rt , xt )0 ,
xt is the dividend yield,
and
t ∼ N(0, Σ)
Lecture 7
Mark-Jan Boes 43
Barberis: analysis 2
Parameter estimates:
Lecture 7
Mark-Jan Boes 44
Barberis: analysis 2
Lecture 7
Mark-Jan Boes 45
Barberis: analysis 2
Lecture 7
Mark-Jan Boes 46
Barberis: analysis 2
Lecture 7
Mark-Jan Boes 47
Asset allocation in an asset-only framework
In a world where returns are i.i.d. and the investor has power
utility, the optimal asset allocation would be identical through time:
the myopic investor would rebalance regularly to this asset mix.
Lecture 7
Mark-Jan Boes 48
Asset allocation in an asset-only framework
In a world where returns are not i.i.d., the myopic investor would
update expected returns, variances and covariances at the
rebalancing date and would determine the new optimal asset mix.
Long term investors should act like short term investors but, in
addition to that, also have so-called ’hedging demands’.
Lecture 7
Mark-Jan Boes 50
Barberis: analysis 3
Lecture 7
Mark-Jan Boes 51
Barberis: analysis 3
Lecture 7
Mark-Jan Boes 52
Barberis: analysis 3
Lecture 7
Mark-Jan Boes 53
Asset allocation
Ang splits the long run weight at a particular point in time in:
short run weight
opportunistic weight
The short run weight is the weight a short term investor would
choose; it can vary through time because expected returns and
variance vary through time.
He calls the long run weights, the strategic asset allocation, the
portfolio allocation a long term investor really wants to have.
Lecture 7
Mark-Jan Boes 54
Asset Allocation
Lecture 7
Mark-Jan Boes 55
Asset Allocation
Lecture 7
Mark-Jan Boes 56
Higher moments
’fat left tails’, i.e. occurs when extreme negative returns are
observed with a magnitude and frequency greater than implied
by the ’normal’ distribution.
correlation breakdown in joint asset class returns, i.e. occurs
during periods of high market volatility and is typically not
captured by linear correlation matrices
Lecture 7
Mark-Jan Boes 57
Higher moments
Lecture 7
Mark-Jan Boes 58
Higher moments
Lecture 7
Mark-Jan Boes 59
Higher moments
Lecture 7
Mark-Jan Boes 60
Higher moments
On the next slide I’ll presents something that looks very familiar to
a mean-variance efficient frontier.
Lecture 7
Mark-Jan Boes 61
Higher moments
Lecture 7
Mark-Jan Boes 62
Higher moments
Lecture 7
Mark-Jan Boes 63
HigherEmoments
xhibit 18
Optimization Results
this article, forward to an unauthorized user or to post electronically without Publisher permission.
tments 2009.12.3:8-35. Downloaded from www.iijournals.com by CAIA on 05/21/10.
Lecture 7
Mark-Jan Boes 65
Summary
What do we learn from all this and could be of value for the
pension fund portfolio choice problem?
In general, portfolio choice is about:
modelling the world (i.e. state variables and asset classes)
over the full investment period
determining the utility function: which goals and which risks
are important?
choosing the type of investor: buy-and-hold, myopic or
dynamic investor?
maximizing expected utility
Lecture 7
Mark-Jan Boes 66
Summary
Lecture 7
Mark-Jan Boes 67
Summary
Lecture 7
Mark-Jan Boes 68