IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

IB9Y8 Asset Pricing

Problem Set 2-1: Solution Notes

Question 1: We discuss each statement in turn:

a) Without the investment project, the investor would most likely not borrow today
but put part of their income on deposit (“save”) to increase future consumption.

TRUE: From the shape of the indifference curve shown, it is obvious that the point of
tangency with the “without-project” opportunity set must lie to the “north-west”
of the income point (note that indifference curves can never intersect).

b) The investor can borrow at a risk-free interest rate of approximately 5%, and also
earns 5% interest on risk-free deposits (“savings”).

TRUE: The line depicting the opportunity set goes from 100,000 current consumption to
105,000 future consumption (without the project), that corresponds to a slope of
–1.05, and hence an interest rate of 5%.

c) To undertake the investment project and implement the optimal consumption

plan, the investor would have to borrow £40,000 today.

FALSE: Although the “income with” point lies at 30,000, and the “income without” point
at 70,000, which implies that the project requires 40,000 investment now, not all
of this must be borrowed. We only need to borrow the difference between the
income after investing in the project (30,000) and the desired optimal current
consumption (around 55,000).

d) The net present value (“NPV”) of the project is approximately £10,000. Since this
is clearly positive, the investor will undertake the project.

TRUE: The difference on the horizontal axis (“consumption today”) between the two
opportunity sets “with” and “without” is, according to the picture, just a little
over 10,000, and this must hence be the project’s NPV.

1
Question 2: Note that u ( c ) = ln ( c ) implies u′ ( c ) =
.
c
Y 100,000
a) Maximum current consumption = Y0 + 1 = 100,000 + = 195,238 .
1+R 1.05

b) From the lectures, we know that the first-order condition for optimality is:
1+R
u′ ( C 0 ) = u′ ( C1 ) .
1+ β
1
With β = 0 and u′ ( c ) = , this implies: C1 = (1 + R ) C 0 .
c

Page 1 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Q2 (cont’d): By definition, C 0 = Y0 − δ and C1 = Y1 + (1 + R ) δ . We thus get:

1 Y 
Y1 + (1 + R ) δ = (1 + R )(Y0 − δ ) , which we can solve: δ =  Y0 − 1  .
2 1+R 

Substituting the optimal δ back into consumption, we find:

1 Y 
C 0 =  Y0 + 1  , and from above C1 = (1 + R ) C 0 .
2 1+R 

Using the numbers we are given, we calculate:

1 100,000 
C 0 =  100,000 +  = 97,619 , and hence C1 = 102,500.
2 1.05 

In other words, we put 2,381 on deposit today, and next period collect the return
2,500 (deposit plus 5% interest) to increase future consumption to 102,500.

c) Without any calculations, it is obvious that the project has a return > 5%, so it
must have a positive NPV. And indeed:

85,000
NPV = − 80,000 = 952 (to the nearest £1).
1.05

d) Residual income today after investing in the project is Y0 – 80,000 = 20,000, and
gross income next period (including the project pay-off) is Y1 + 85,000 = 185,000.
Substituting this into the calculations from b), we get:

1 185,000 
C 0 =  20,000 +  = 98,095 , and C1 = 103,000 (nearest £1).
2 1.05 

Note that optimal current consumption has increased by 476 = 952/2, exactly one
half of the project’s NPV (as the equation for optimal consumption suggests: the
optimal consumption today is half of the present value of all income).

The plan is thus: take 80,000 of our 100,000 current income and invest. To fund
the desired consumption, we must hence borrow 78,095. Next year, we pay back
1.05 × 78,095 = 82,000 to settle our debt. Together with income and the pay-off
from the project, we have 100,000 + 85,000 – 82,000 = 103,000 left.

e) Of course, by Fisher separation, our decision to undertake the project remains.

However, next year’s income (including project pay-off) is only 125,000. Hence:

1 125,000 
C 0 =  20,000 +  = 69,524 , and C1 = 73,000.
2 1.05 

In other words, it is not the level of current consumption that the optimal plan
targets, but the smooth distribution of wealth across time.

Page 2 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Question 3: We start again with the FOC as in 2.b). With u′ ( c ) = c − γ , we get:

1
1 + R −γ  1 + R γ
C 0−γ = C1 , which we can re-arrange into: C1 =   ⋅C0 .
1+ β 1+ β 
 
=: α
Again, using C 0 = Y0 − δ and C1 = Y1 + (1 + R ) δ , we can solve this for δ:

δ=
α ⋅ Y0 − Y1
, which in turn implies: C 0 =
(1 + R ) ⋅Y0 + Y1 .
(1 + R ) + α (1 + R ) + α
1
 1 + R γ
b) For β = 0 and γ = 2, we get α =   = 1.05 = 1.0247. Hence:
1+ β 

1.05 × 20,000 + 125,000

C0 = = 70,372 , and C1 = α ⋅ C0 = 72,110 .
1.05 + 1.0247

Compared to the solution in 2.e), we have slightly increased current consumption

at the expense of next year’s. Note however that:

73,000 72,110
69,524 + = 139,048 = 70,372 +
1.05 1.05

The aggregate present value of current and future consumption is the same here
as it was in Question 2.e) above. We have not created or “lost” any real value,
just allocated the total differently across time.

Question 4: As the footnote in the question paper suggests, we use calculus of variations. We
assume that we have calculated and implemented the optimal plan, including the
investment θ in the risky asset. Now suppose we deviate from the optimum by
buying a tiny amount ε more of the asset. This reduces current consumption by
an amount ε⋅P0 and increases future consumption by ε⋅X1.

Now here’s the argument: because the original allocation was optimal, if we now
evaluate the utility function along different values of ε, it must have a maximum
at the point ε = 0. So, we’re back to first-order conditions:

d  1  1
0= u ( C 0 − ε P0 ) + E u ( C1 + ε X1 )   = −P0u′ ( C 0 ) + E u′ ( C1 ) ⋅ X1 
dε ε =0  1+ β  1+ β 

Dividing by u′ ( C 0 ) , and taking P0 to the left-hand side:

  1 u′ ( C 1 )  
P0 = E   ⋅  ⋅ X 1  , as claimed.
1 + β u (C 0 ) 
  ′ 

=: M

Page 3 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Q4 (cont’d): Thus, we can indeed write P0 = E [M ⋅ X1 ] .

This is, in a way, our first real “asset pricing equation”. What it says is: the price of
the asset is given by the expectation of the asset’s pay-off, multiplied by M. Note
that M is stochastic: C1 is a function of income Y1 and asset pay-off X1, both of
which are uncertain ex-ante.

Much like a probability density function, we can think of M as assigning “weights”

to different “states of nature” which describe how valuable a unit of pay-off is to
the investor in that particular state. For this reason, we call M the “Stochastic
Discount Factor” that prices the assets in our model.

b) Using the definition of covariance, we re-write the pricing equation as:

P0 = E [M ⋅ X1 ] = cov ( M , X1 ) + E [M ] E [ X1 ]

In other words, the asset price P0 is positively related to cov( M, X1 ). But from
the definition of M and the concavity of u (i.e. u’’ < 0), we see that M is inversely
correlated with C1 and thus also with Y1.

This has a nice intuitive interpretation: the asset has a higher value if its pay-offs
are negatively correlated with income. Put differently, we value an asset more if
it pays us a lot in times when we are poor (i.e. when we need it most).

The propensity to smooth consumption not only applies across time, but also
across different future “states of the world”. Part of a financial asset’s value thus
derives from its usefulness as a hedge against income uncertainty.

Question 5: Clearly, for each roll there are six possible outcomes, {1, 2, 3, 4, 5, 6}. As all of
these outcomes are equally likely, the probability of any one occurring is p = 1/6.

a) Trivial: the probability of a “6” in one roll is p = 1/6.

b) Also trivial: the probability is still 1/6, see “Note” below.

c) Here, we use a little “probabilistic algebra”: denote by Xk the outcome of the k’th
roll. Thus defined, each Xk is a random variable, i.e. a variable with a value that is
not known ex-ante, but for which the distribution of possible outcomes is known.

Note: For the case considered here, the random variables X1, X2, … are independent: at
the time we make the k’th roll, we do not need to know what outcomes we have
observed in the preceding k – 1 rolls. Even if there has not been a single “6” in 99
rolls, the probability of scoring a “6” in the 100’th roll is still 1/6. The dice has “no
memory”: when you make the 100’th roll, the dice “does not remember” the fact
that it has not shown you a “6” in the past 99 rolls!
This is a mistake often made by gamblers: for example in Roulette, many
would argue like this: “in the last 10 rounds, the ball always landed on ‘red’; it
just has to land on ‘black’ this time!” This is nonsense: the chance of “black” in
the next round is still 50% (well, to be precise it’s 18/37 = 48.6% because of the
“zero” on the roulette wheel). Again, the reason is that neither the ball nor the
wheel has any “memory” of the fact that there were 10 “reds” before …

Page 4 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Q5 (cont’d): For independent random variables, probability is multiplicative:

N
1
prob {{ X1 = 6} ∧ … ∧ { X N = 6}} = ∏ prob{ X k = 6} = N
k =1

  6
=1/6

Here, the symbol “∧” stands for “and”. In other words, the left-hand side above
reads “the probability that X1 = 6 and X2 = 6 and … and XN = 6”. The equation thus
states that this probability is the product of the probability (= 1/6) for each Xk = 6.

c (i) Using the above, the probability of observing 3 outcomes of “6” in 3 rolls of the
dice is 1/63 = 1/216 = 0.0046296 = 0.46296%.

c (ii) I wasn’t very precise when I wrote this question: did I mean “observe exactly 3
outcomes of ‘6’ in 100 rolls”, or “at least 3 outcomes …”? So I’ll do both variants:

 100 
For “exactly 3 …” we use Appendix A. We have:   = 161,700 . Hence:
 3 

prob{exactly 3 outcomes of "6" in 100 rolls}

100
 5  100  −3 161,700
 ⋅ 5 = (1.2075 × 10 ) ⋅
−8
=  ⋅ = 1.5620 × 10−5 = 0.001562%
6  3  125

For “at least 3 …”, we use Appendix B:

 100   100   100 

With   =1,   = 100 , and   = 4,950 , we get:
 0   1   2 

prob {at least 3 outcomes of "6" in 100 rolls}

100
 5 2
 100  − k
=1−  ⋅∑  ⋅5
6 k =0  k 

 1 100 4,950 
= 1 − (1.2074 × 10−8 ) ×  + + −6
 = 1 − 2.6444 × 10 = 99.999736%
1 5 25 

Now that we know how it is done, we just summarize the results here:

Probability
obs. # of ‘6’s … in # of rolls exactly … at least …
d, e (i) 1 3 34.72% 42.13%
d, e (ii) 1 100 0.000024% 99.999998%
c (i) 3 3 0.46296% 0.46296%
c (ii) 3 100 0.001562% 99.999736%

Page 5 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Notes: Our experiments show a few things:

Even with just a few rolls, there is a rather high chance of observing at least
one outcome of a given value (e.g. 42% for only 3 rolls). This likelihood increases
as we have more rolls (for just 4 rolls, we already have a more than 50% chance of
scoring at least one “6”); in the limit (as the number of rolls increases) this
likelihood approaches 100%.
While the likelihood of observing any given (small) pattern of outcomes at
least once approaches 100% as we increase the number of rolls (for 45 rolls, we
already have a 99% chance of observing any three-value combination at least
once), the likelihood of observing exactly one occurrence of a given pattern tends
to zero (for 55 or more rolls there is a less than 1% chance of observing exactly
one occurrence of any combination of 3 values).

Important: While we used a simple experiment of rolling dice to derive the above results,
there are implications for the world of finance! Assume that consecutive returns
on financial assets are serially un-correlated. This is not as strong an assumption
as it may seem: it is in fact an implication of the paradigm of “market efficiency”
(see Chapter 3). In this case, our results translate such:

If we look at the time series of asset returns, we are practically certain to find
any given pattern we care to find, if only we keep looking long enough! But this
tells us precisely nothing about the returns that will follow the pattern!

What does this say about those people who pay exorbitant fees to managers who
run investment funds based on “technical analysis1”?

1
“Technical Analysis” refers to the “art” (?) of spotting patterns (e.g. support and resistance levels, “head and
shoulders”, and many others; look these up if you care) and using these to “predict” future returns.

Page 6 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Question 6: Here, we use a little Bayesian probability calculus. Denote by “D+” and “D–“ the
events that a given test subject “does (+) or does not (–) have the disease”. And
denote by “T+” and “T–” the test outcome (positive or negative). Following
standard notation, we write “P(A|B)” to mean the probability of A conditional on
knowing that B is true.

As Louise has tested positive, we know “T+” in her case. We want to find …

P ( D+ ∧ T+ )
P ( D+ |T+ ) =
P (T+ )
P (T+ |D+ ) P ( D+ )
=
P (T+ )
P (T+ |D+ ) P ( D+ ) 99% × 1% 1
= = =
P ( T+ |D+ ) P ( D+ ) + P ( T+ |D− ) P ( D− ) 99% × 1% + 1% × 99% 2

What is going on here?

The result seems counter-intuitive: although the test is super-accurate (getting it

“wrong” in only 1% of cases), a positive test result gives only a 50/50 indication
that Louise carries the virus! But it’s really simple:

Suppose we test 10,000 people randomly picked. There will be 100 who carry the
virus, and 9,900 who are “clean”. Now we test all. Of the 100 with virus, we’ll
get 99 positive tests, and one (incorrect) negative. Of the 9,900 clean subjects,
we’ll get 99 (1%) of (incorrect) positive results, and 9,801 with (correct) negative
result. Now concentrate on those that tested positive (in our algebra above, this
means “conditioning” on T+). In this group, there are 99 with, and 99 without the
virus! As we had picked randomly, we are equally likely that our subject in
question (Louise) is in either of these two groups, i.e. a 50/50 chance.

Page 7 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Appendix A: Observing exactly k outcomes of a given value in N draws

We use an approach known as “combinatorics”: consider a random experiment

with a multitude of different possible outcomes, all of which are equally likely. To
compute the probability of observing an outcome that has a certain property, we
count the number of possible outcomes that possess the property, and divide by
the number of all possible outcomes. For example, when rolling a dice, there are
6 possible outcomes, and of these 3 are prime numbers. Hence, the probability of
observing a prime in one roll of a dice is 3/6 = 50%.

Suppose we have a random variable Xk that can take on the values 1, 2 … M, each
with equal probability p = 1/M. Fix any number x in the range 1 ≤ x ≤ M.

We ask: If we draw N independent outcomes X1 … XN, what is the probability of observing

exactly k ≤ N outcomes of value “x”?

Step 1: The k outcomes of “x” can be in different places in the sequence X1 … XN. The
number of ways of selecting k elements from a sequence of N is2:

N N! ( N − k + 1 ) ⋅ ( N − k + 2 ) ⋅… ⋅ ( N − 1 ) ⋅ N
 = =
 k  k ! ⋅ ( N − k )! 1 ⋅ 2 ⋅… ⋅ ( k − 1 ) ⋅ k

Let’s check this: how many ways are there to arrange 2 occurrences of “x” in a
sequence of 3 outcomes? There are three: (x, x, ?), (x, ?, x), and (?, x, x). Indeed,
for this example we have N = 3 and k = 2, and we get:

 3 2⋅3
 = =3
 2  1⋅2

Step 2: For each choice of k positions in a sequence of N elements, how many sequences
are there that have “x” in these positions and “non-x” in all other positions? In
each of the (N – k) “other” positions, there can be any one of (M – 1) possible
(“non-x”) values. Thus, there are (M – 1)N – k sequences with “x” in the k chosen
positions and values different from “x” in all of the other N – k positions.

Step 3: Finally, how many possible outcomes are there in total in our experiment? In
each of the N positions of the sequence of draws, there can be any one of M
values. Thus, there are in total MN possible sequences. Putting it all together:

Answer: prob{exactly k outcomes of "x" in N draws}

N
1 N   M −1  N 
= N ⋅   ⋅ ( M − 1) =  ⋅   ⋅ ( M − 1)
N −k −k

M k  M  k

For example, the likelihood of observing exactly two outcomes of “6” in 3 rolls of
a dice is: (5/6)3 × 3 / 52 = 0.0694 = 6.94%. To check: we know there are 3 ways of
arranging two “6”s in 3 outcomes, and for each there are 5 values to go in the
“non-6” space. And (3 × 5) / 63 = 15 / 216 = 0.0694.

2
These are known as the “binomial coefficients”, or “coefficients of choice”; look this up for more.

Page 8 of 9
 IB9Y8 Asset Pricing Problem Set #2-1: Solution Notes

Appendix B: Observing at least n ≤ N outcomes of a given value in n draws

We will use the result from Appendix A. Note first:

{at least n outcomes of "x" in N draws}

N
= ∪ {exactly k outcomes of "x" in N draws}
k=n

Here, the symbol “U” means the “union” of events: the right-hand side in the
above equation reads: “the event ‘observe exactly k …’ is true for at least one of
the numbers k = n … N.”

Next we observe that the events on the right-hand side (“observe exactly k …”)
are mutually exclusive for different k. Hence, the probability is additive when we
form the union of such events:

prob{at least n outcomes of "x" in N draws}

N
1 N N 
= ∑ prob {exactly k outcomes …} = N ∑   ⋅ ( M − 1)
N −k
⋅
k =n M k =n  k 

In the last step, we have used the result from Appendix A. Now note that:

N
N  N
N  n −1 N
 
∑  k  ⋅ ( M − 1) = ∑   ⋅ ( M − 1) − ∑   ⋅ ( M − 1 )
N −k N −k N −k

  k=0  k  k
k=n
  k =0  
= (1 + M − 1) = M N
N

N
N 
The last step follows from3: ( x + y ) = ∑   ⋅ x k ⋅ y N − k with x = 1 and y = M – 1.
N

k =0  k 

Result: Putting everything together:

prob{at least n outcomes of "x" in N draws}

N
 M − 1  n −1  N 
 ⋅ ∑  k  ⋅ ( M − 1)
−k
=1−
 M  k =0  

Note: The last expression, by the laws of probability, implies:

prob{fewer than n outcomes of "x" in N draws}

N
 M − 1  n −1  N 
 ⋅ ∑  k  ⋅ ( M − 1)
−k
=
 M  k =0  

3
This is the well-known “binomial formula”.

Page 9 of 9