ECO 503 Problem Set 3 Solutions
1. For u (x) = ax + b where a > 0 and b → R, for any lottery L2 and L1 with distributions F2 and F1 :
L2 ↭ L1 ↑↓ U (L2 ) ↔ U (L1 )
↑↓ EF2 [u (x)] ↔ EF1 [u (x)]
↑↓ EF2 [ax + b] ↔ EF1 [ax + b]
↑↓ aEF2 [x] + b ↔ aEF1 [x] + b
↑↓ EF2 [x] ↔ EF1 [x]
↑↓ E [L2 ] ↔ E [L1 ]
Here, the notation EFi [x] means the expected value is computed using the distribution Fi (·).
2. In this lottery, there are six equally likely outcome, and in two of them Mobin gets +18, and in the remaining
four he gets ↗9.
!2 " ! "
(a) The expected outcome of money received with the die-throwing lottery G is E [G] = 6 ↘ 18 + 46 ↘ (↗9) =
0. Therefore, this is a fair bet.
! " ! " ≃ ≃ ≃ ≃
(b) Here, U (G) = E [u (G)] = 13 ↘ u (9 + 18) + 23 ↘ u (9 ↗ 9) = 13 27 + 23 0 = 13 ↘ 3 3 = 3. On the
≃ ≃
other hand, U (G0 ) = 1 ↘ u (9) = 9 = 3. Because 3 > 3, we have U (G0 ) > U (G), so for Mobin,
G0 ⇐ G, and he will reject the die-tossing gamble.
(c) We are now looking for the smallest w for which U (G) ↔ U (G0 ). As before, U (G0 ) = 3. However, if w
! ≃ " ! ≃ " ≃
is the amount from winning, we will have U (G) = 13 ↘ 9 + w + 23 ↘ 0 = 13 9 + w. We need:
1≃
9+w ↔ 3
3
≃
9+w ↔ 9
9+w ↔ 81
w ↔ 72
Which means, w = $72 is the lowest winning amount for which he will accept the lottery G.
1
3. Stock A: $50 with probability 12 , $250 with probability 12 .
Stock B: $100 with probability 12 , $200 with probability 12 .
(a) The expected profits are:
$ # $#
1 1
E [A] = ↘ 50 + ↘ 250 = 150
2 2
# $ # $
1 1
E [B] = ↘ 100 + ↘ 200 = 150
2 2
So the expected profits are equal.
(b) Now, we have to compute the expected utilities of choosing A or B.
≃ ≃
U (A) = 12 ↘ 50 + 12 ↘ 250 ⇒ 11.44
≃ ≃
U (B) = 12 ↘ 100 + 12 ↘ 200 ⇒ 12.07
As we can see, U (B) > U (A), so for Mobin, B ⇐ A, and he will choose stock B.
4. Both lotteries are associated with distributions over [0, 1]. f1 (x) = 1 (uniform distribution) and f2 (x) = 2x.
x x
(a) F1 (x) = x and F2 (x) = x2 . For x → [0, 1], x2 ⇑ x, so F2 (x) ⇑ F1 (x), that is, f2 (t)dt ⇑ f1 (t)dt
0 0
1
(b) U (Li ) = u(x)fi (x)dx
0
i. When u(x) = x:
1 % &1
x2
U (L1 ) = xdx = 2 = 12 ;
0 0
1 % &1
2x3 2
U (L2 ) = x(2x)dx = 3 = 3
0 0
ii. When u(x) = x2 :
1 % &1
x3 1
U (L1 ) = x2 dx = 3 = 3 ⇒ 0.3333;
0 0
1 % &1
2x4 1
U (L2 ) = x2 (2x)dx = 4 = 2
0 0
1
iii. When u(x) = x 2 :
1
1
% 3 &1
2
U (L1 ) = x 2 dx = x2
3 = 3 ⇒ 0.6667;
0 2 0
1
1
% 5 &1
2x 2 4
U (L2 ) = x 2 (2x)dx = 5 = 5
0 2 0
(c) For all of the three utility functions, the DM would prefer L2 over L1 . L1 puts the same probability on
all outcomes over [0, 1], whereas L2 puts higher probabilities on higher outcomes. Seems like if the DM
prefers higher outcomes (that is, if her utility function is increasing), then she would prefer L2 over L1 .
(d) I considered long and hard the lottery L3 and its cdf:
1
2x2 if 0 ⇑ x ⇑ 2
F3 (x) =
1
4x ↗ 2x2 ↗ 1 if <x⇑1
2
2
x x
One interesting fact one might observe is that for any x → [0, 1], F3 (t)dt ⇑ F1 (t)dt.
0 0
(e) The piece-wise defined pdf is:
1
4x if 0 ⇑ x ⇑ 2
f3 (x) =
1
4 ↗ 4x if <x⇑1
2
(f) The expected utilities under L3 are:
i. When u(x) = x:
1
2 1 % & 12 % &1
4x3 4x2 4x3 1 1 1
U (L3 ) = x(4x)dx + x(4 ↗ 4x)dx = 3 + 2 ↗ 3 1
= 6 + 3 = 2
0 1 0 2
2
ii. When u(x) = x2 :
1
2 1 + ,1 % 3 &1
U (L3 ) = x2 (4x)dx + x2 (4 ↗ 4x)dx = x4 02 + 4x3 ↗ x4 1 = 16
1
+ 11
48 = 7
24 ⇒ 0.2917
0 1 2
2
1
iii. When u(x) = x 2 :
1
2
1
1
1
% 5 & 12 % 3 5 &1 → → →
4x 2 4x 2 4x 2 2 16 7 2 16↑4 2
U (L3 ) = x 2 (4x)dx+ x 2 (4↗4x)dx = 5 + 3 ↗ 5
1
= 5 + 15 ↗ 15 = 15 ⇒ 0.6895
0 1 2 0 2 2
2
2
(g) The DM’s preferences over the three lotteries:
i. When u(x) = x: L2 ⇐ L1 ⇓ L3
ii. When u(x) = x2 : L2 ⇐ L1 ⇐ L3
1
iii. When u(x) = x 2 : L2 ⇐ L3 ⇐ L1
(h) We can see that when u(x) = x (linear), the DM is indi!erent between L3 and L1 . When u(x) = x2
1
(convex) the DM prefers L1 over L3 . When u(x) = x 2 (concave), the DM prefers L3 over L1 .
1
Both L1 and L3 have the same expected outcome of 2. But whereas L1 puts equal probability on all
outcomes, L3 puts lower probabilities on extreme outcomes (close to 0 and 1), and higher probabilities
on outcomes close to the average. If the DM has increasing marginal utility, she prefers the more risky
lottery L1 , if the DM has decreasing marginal utility, she prefers the less risky lottery L3 , and if she has
constant marginal utility, she’s indi!erent (risk-neutral).
≃
5. Here, u(x) = x, wealth level is w = 5.
(a) The A-P measure of absolute risk aversion is
3
u↓↓ (w) ↗ 1 w↑ 2 1 1
r(w) = ↗ ↓ = ↗ 14 ↑ 1 = w↑1 =
u (w) 2w
2 2 10
3
! "
(b) For the lottery L = 16, 4; 12 , 12 , the certainty equivalent c solves:
u(c) = U (L)
≃ 1≃ 1≃
c = 16 + 4
2 2
≃ 1 1
c = (4) + (2)
2 2
≃
c = 3
↫c = 9
! "
For the lottery L↓ = 36, 16; 12 , 12 , the certainty equivalent c↓ solves:
u(c↓ ) = E [u(L↓ )]
≃ 1≃ 1≃
c↓ = 36 + 16
2 2
≃ 1 1
c↓ = (6) + (4)
2 2
≃
c↓ = 5
↫ c↓ = 25
6. If v (x) = f (u (x)), where v, u, f are all strictly increasing and strictly concave functions, then, the A-P
coe"cient for the v utility function is:
v (x)
→→
rv (x) = ↗
v → (x)
And the A-P coe"cient for the u utility function is:
u (x)
→→
ru (x) = ↗
u→ (x)
4
Now, notice that:
v (x) = f (u (x))
↫ v (x) = f (u (x)) u (x)
→ → →
% → &2
v (x) = f (u (x)) u (x) + f (u (x)) u (x)
→→ →→ → →→
v (x)
→→
now, rv (x) = ↗
v → (x)
% → &2
f (u (x)) u (x) + f (u (x)) u (x)
→→ → →→
= ↗
f → (u (x)) u→ (x)
- .
f (u (x)) u (x) u (x)
→→ → →→
= ↗ + →
f → (u (x)) u (x)
- →→ .
u (x) u (x)
→
= ↗ f (u (x)) → + ↗ →
→→
/ f (u (x)) u (x)
<0 /
>0
/
>0
> ru (x)
↫ rv (x) > ru (x)
7. (Optional to study) Asset 1 gives outcome of 1 for sure, and asset 2 gives outcomes a and b with probabilities
ω and 1 ↗ ω, respectively.
(a) If min {a, b} ↔ 1, then the risky asset always gives a higher return. Therefore a simple necessary condition
for the demand for the risk-less asset to be strictly positive is min {a, b} < 1.
(b) If the expected outcome of the risky asset is less than or equal to 1, then the risk-averse DM would never
choose it1 . So a simple necessary condition for the demand for the risky asset to be strictly positive is
ωa + (1 ↗ ω)b > 1
(c) Since the prices of both assets are equal to 1, at the optimally chosen quantities (x1 , x2 ), the (expected)
marginal utilities of the two assets must be equal. The DM’s expected utility is given by ωu (x1 + x2 a) +
(1 ↗ ω)u (x1 + x2 b). Henceforth,
ωu↓ (x1 + x2 a) + (1 ↗ ω)u↓ (x1 + x2 b) = ωau↓ (x1 + x2 a) + (1 ↗ ω)bu↓ (x1 + x2 b)
1 note this would not necessarily be true if the DM is risk-loving.
5
Simplifying, we get
↫ ω(1 ↗ a)u↓ (x1 + x2 a) + (1 ↗ ω)(1 ↗ b)u↓ (x1 + x2 b) = 0
Plugging in the budget constraint x1 + x2 = 1, gives us the necessary first order condition:
ω(1 ↗ a)u↓ (x1 + (1 ↗ x1 )a) + (1 ↗ ω)(1 ↗ b)u↓ (x1 + (1 ↗ x1 )b) = 0
(d) Taking b as given, let us define
ε(a, ω, x1 ) = ω(1 ↗ a)u↓ (x1 + (1 ↗ x1 )a) + (1 ↗ ω)(1 ↗ b)u↓ (x1 + (1 ↗ x1 )b)
Remember that the demand for the risky asset, x1 , is the DM’s choice, and should be thought of as a
function of a, x1 (a). Based on the first order condition, ε (a, ω, x1 (a)) = 0 (constant) for any value of a.
This means, for a fixed ω,
ϑε ϑε dx1
ε (a) = + =0
→
ϑa ϑx1 da
Now,
ϑε
= ↗ωu↓ (x1 + (1 ↗ x1 )a) + ω(1 ↗ a)u↓↓ (x1 + (1 ↗ x1 )a) (1 ↗ x1 ) < 0
ϑa
In claiming this, we used the fact that u↓↓ is negative, as the DM is risk-averse.
And,
ϑε
= ω(1 ↗ a)u↓↓ (x1 + (1 ↗ x1 )a) (1 ↗ a) + (1 ↗ ω)(1 ↗ b)u↓↓ (x1 + (1 ↗ x1 )b) (1 ↗ b)
ϑx1
= ω(1 ↗ a)2 u↓↓ (x1 + (1 ↗ x1 )a) + (1 ↗ ω)(1 ↗ b)2 u↓↓ (x1 + (1 ↗ x1 )b) < 0
Now going back to the condition before,
ϑε ϑε dx1
+ =0
ϑa ϑx1 da
/ /
<0 <0
We can see then we must have dx1
da < 0 as well.
(e) Still assuming a < 1, we can say by the condition in part (b) that b > 1, which means a is the worse
outcome. If ω, which is the probability of the worse outcome, goes up, we would think the demand for
the risky asset would go down, and the demand for the risk-less asset, x1 , would go up. So we would
conjecture dx1
dω > 0.
6
(f) It’s pretty much the same kind of exercise as part (d) to check this. This time, taking a < 1 < b as given,
we can think of x1 as a function of ω, x1 (ω). Then,
ϑε ϑε dx1
+ =0
ϑω ϑx1 dω
Now,
ϑε
= (1 ↗ a)u↓ (x1 + (1 ↗ x1 )a) ↗ (1 ↗ b)u↓ (x1 + (1 ↗ x1 )b) > 0
ϑω
Which means,
ϑε ϑε dx1
+ =0
ϑω
/ ϑx1 dω
/
>0 <0
This tells is that dx1
dω > 0, as expected.