Unit 3: Risk and Time Preference

Andrew Caplin and Srijita Ghosh

Fall 2023

1 / 128
Unit 3: Risk and Time Preference

▶ Next 4 lectures on choice under uncertainty

▶ Key to many important financial and insurance markets
▶ 3.1: Lotteries and the Expected Utility Theorem
▶ 3.2: Monetary Prizes and Risk Attitudes
▶ 3.3: Risk Aversion and Insurance
▶ 3.4: Portfolio Allocation
▶ Additional Material

2 / 128
 3.1. Lotteries and the EU Theorem 3.1.1: Lotteries

3.1. Lotteries and the EU Theorem

3.1.1: Lotteries

3 / 128
 3.1. Lotteries and the EU Theorem 3.1.1: Lotteries

3.1.1 Lotteries

▶ New choice domain of lotteries over a finite set N of prizes.

▶ Let Z be the set of prizes (winning a car, $ 1000 etc.)
▶ Choice set is not Z but X
▶ X set of lotteries over prizes in Z defined as follows:
N
X = {p = (p1 , ..., pN ) ∈ RN
+| ∑ pn = 1}.
n =1

▶ We are given a complete and transitive binary relation ≿ over

X comparing lotteries according to personal likes and dislikes.

4 / 128
 3.1. Lotteries and the EU Theorem 3.1.1: Lotteries

3.1.1 Lotteries

▶ In the case of lotteries, there is a very important interpretation

of the weighted average, which is in terms of a compound lottery
▶ Given any two lotteries p, q ∈ X and λ ∈ [0, 1] , there exists
one and only one corresponding simple lottery, which we denote,

λp + (1 − λ)q ≡ (λp1 + (1 − λ)q1 , ..., λpN + (1 − λ)qN ) ∈ X .

▶ Simple lottery describes final prizes and corresponding

probabilities

5 / 128
 3.1. Lotteries and the EU Theorem 3.1.1: Lotteries

3.1.1 Lotteries

▶ The explicit tree diagram for the two prize case, with z1 left
branch, z2 right branch.

Figure: tree diagram for two prize case

6 / 128
 3.1. Lotteries and the EU Theorem 3.1.1: Lotteries

3.1.1 Lotteries
▶ There are many different representations of lotteries based on
compounding
▶ Illustrate for (0.5, 0.5) either as a direct lottery or broken up as
0.5(p1 , 1 − p1 ) + 0.5(1 − p1 , p1 ) , p1 ∈ [0, 12 ] .

Figure: Left: simple lottery, Right: compound lottery

7 / 128
 3.1. Lotteries and the EU Theorem 3.1.1: Lotteries

3.1.1 Lotteries

▶ By defining preferences on the lottery domain we insist that all

distinct resolution devices are indifferent
▶ all that matters is final prize probabilities
▶ Work with corresponding unique simple lottery
▶ Lotteries suggest possible structure to preferences

8 / 128
3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2: EU Theorem

9 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

▶ A technical assumption of continuity ensures existence of a

utility function that represents the preference over lotteries.
▶ Continuity (A1): Given p, q, r ∈ X such that p ≻ q ≻ r there
exists λ ∈ (0, 1) such that,

λp + (1 − λ)r ∼ q.

▶ A complete and transitive binary relation that satisfies A1 has a

utility function representation

10 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

▶ Two prizes, z1 , z2 with

▶ Suppose z1 ≻ z2 , i.e., you prefer prize z1 over prize z2
▶ What can we say about lotteries on {z1 , z2 }?
▶ Might expect
(0.75, 0.25) ≻ (0.25, 0.75).
▶ Or, more generally

(p, 1 − p ) ≻ (q, 1 − q ) for p > q

11 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

▶ Follows from a more general assumption

▶ Substitution Axiom(A2) : Given p, q, r ∈ X with p ≻ q and
given λ ∈ (0, 1] ,

λp + (1 − λ)r ≻ λq + (1 − λ)r, :

and with p ∼ q given λ ∈ [0, 1] ,

λp + (1 − λ)r ∼ λq + (1 − λ)r.

12 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

▶ Substitution axiom does make intuitive preference true

▶ Reason is that can write both as compound lotteries differing
only in pure prize component (see figure next slide):

(0.75, 0.25) = 0.5(0.5, 0.5) + 0.5(1, 0);

(0.25, 0.75) = 0.5(0.5, 0.5) + 0.5(0, 1).

13 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

Figure: substitution axiom: Up (0.75, 0.25), Down (0.25, 0.75)

14 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

▶ Hence by substitution,

z1 = (1, 0) ≻ (0, 1) = z2 =⇒ (0.75, 0.25) ≻ (0.25, 0.75).

▶ This is general: given two prizes, z1 , z2 with z1 ≻ z2 one with

higher probability of better prize is preferred when the
substitution axiom holds.
▶ If the two prizes are indifferent, so are all lotteries between them

15 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

▶ Expected Utility Theorem: With A1 and A2 there exists a

function u : Z −→ R such that, for any two lotteries p, q ∈ X ,
N N
p ≿ q if and only if ∑ pn u (zn ) ≥ ∑ qn u (zn ).
n =1 n =1

▶ Axioms A1 and A2 imply existence of such a function

u : Z −→ R.
▶ Such u (.) called vNM utility function
▶ A proof is included at the end: not for test
▶ Likewise some critiques

16 / 128
 3.1. Lotteries and the EU Theorem 3.1.2: EU Theorem

3.1.2 EU Theorem

▶ Justification for the name of the theorem and the use of Eu (p)
to denote the expected utility of the lottery p given the utility
function u : R+ −→ R .
▶ Explain expected height as sum of heights divided by number of
people. When many in some height category weight by
proportion
▶ For an arbitrary lottery p ∈ X , Eu (p) denotes the
mathematical expectation of the utility function u : Z −→ R
using those probability weights on prizes assigned by p ,

Eu (p) = ∑ pn u (zn ),

17 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.1: Monetary prizes

3.2 Monetary Prizes and Risk Preference

3.2.1: Monetary prizes

18 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.1: Monetary prizes

3.2.1 Monetary Prizes

▶ Consider case all prizes are monetary

▶ Let R+ be all such prizes
▶ Call the prize c ≥ 0 for final level of consumption.
▶ Space X comprises simple lotteries with a finite number of
prizes.

19 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.1: Monetary prizes

3.2.1 Monetary Prizes

▶ Identify dollar lottery by pairs of dollar amounts and their

probabilities
▶ [(c1 ; p1 ), (c2 ; (1 − p1 ))] offers $c1 with probability p1 , $c2
with probability 1 − p1 .
▶ Can list any finite number of possible such prizes to define
simple lottery
▶ EU theorem implies ranking over all such monetary lotteries
conveyed by EU function,

u : R+ −→ R.

20 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.1: Monetary prizes

3.2.1 Monetary Prizes

▶ There is a wonderful picture of the EU function over money.

▶ Obvious assumption is that u : R+ −→ R is strictly increasing
in c .
▶ Draw an increasing curve and derive rankings over the following
simple lotteries: (next slide)
▶ L1 : $100 for sure.
▶ L2 : $200 with probability 0.5, $50 with probability 0.5
▶ L3 : $200 with probability 0.25, $150 with probability 0.5, $50
with probability 0.25.

21 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.1: Monetary prizes

3.2.1 Monetary Prizes

Figure: Black: EU (L1 ), Red: EU (L2 ), Blue: EU (L3 )

22 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.1: Monetary prizes

3.2.1 Monetary Prizes

▶ The general connection between chords and lotteries, with the

proportionate distance along the chord denoting probability.
▶ Sure prize: point on utility function (black dashed line)
▶ lottery between two prizes: lies on the chord (red dashed line)
▶ lottery between three prizes: lies on the cord joining third point
with the other point on the chord between first and second
point (blue dashed line)

23 / 128
3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2: Risk Preference

24 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2 Risk Preference

▶ Would also like to understand conditions for different risk

attitude. To do this need some additional notation.
▶ E (p) denotes the mathematical expectation of the number of
dollars on offer,
E (p) = ∑ pn cn .
▶ For any lottery p ∈ X , define the certainty equivalent to p is
the unique single dollar prize CE (p ) such that,

Eu (p) = u [CE (p)] .

▶ The geometry of the certainty equivalent.

25 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2 Risk Preference

Figure: certainty equivalent

26 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2 Risk Preference

▶ The expected utility function u : R+ −→ R is risk averse if,
CE (p) ≤ E (p) for all p ∈ X .
▶ The expected utility function u : R+ −→ R is strictly risk
averse if it is risk averse and the inequality is strict all ”true”
gambles (two or more possible prizes).
▶ The expected utility function u : R+ −→ R is risk loving if,
CE (p) ≥ E (p) for all p ∈ X .
▶ The expected utility function u : R+ −→ R is strictly risk loving
if it is risk loving and the inequality is strict all ”true” gambles
(two or more possible prizes).
▶ The expected utility function u : R+ −→ R is risk neutral if,
CE (p) = E (p) for all p ∈ X .

27 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2 Risk Preference

Figure: panel a: risk-averse, panel b: risk-lover,panel c: risk-neutral

28 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2 Risk Preference

▶ Some geometrically clear propositions:

▶ Concavity is equivalent to risk aversion.
▶ Strict concavity is equivalent to strict risk aversion.
▶ Convexity is equivalent to risk loving.
▶ Strict convexity is equivalent to strict risk loving.
▶ Linear is equivalent to risk neutral

29 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2 Risk Preference

▶ Intuitively, risk aversion ties to insurance (use chord)

▶ Intuitively, risk loving ties to gambling (use chord)
▶ Mixed cases are possible (Friedman and Savage in the big,
Kahneman and Tversky in the small).

30 / 128
 3.2 Monetary Prizes and Risk Preference 3.2.2: Risk Preference

3.2.2 Risk Preference

▶ The strictly concave case is the most important and well-studied

▶ Quantitative measures of risk aversion are extremely important
▶ Geometrically this is how quickly the slope changes

31 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3 Insurance
3.3.1 Optimal Choice of Insurance

32 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ Two states of the world with probabilities: prob π for state 1,

1 − π for state 2.
▶ Pure prize set Z comprises contingent consumption levels, c1
in state 1, c2 in state 2.
▶ Illustration of EU preferences in R2+ (the c1 − c2 plane)
▶ Continuity and Substitution axiom assumed to hold.
▶ EU Theorem: vNM utility function (u (.)) for consumption in
each state is independent of other states (additively separable)
▶ Insurance: ensures sufficient consumption in bad state, i.e.,
where loss is incurred

33 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

Figure: EU preferences for concave u (c ) given π

34 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ EU function formally written as,

EU = πu (c1 ) + (1 − π )u (c2 ).

▶ Note that strict monotonicity as standard for NE direction of

improvement
▶ Risk aversion implies standard shape, risk-neutral linear, risk
loving wrong shape

35 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ With general probabilities get derivative along IC:

πu ′ (c1 ) + (1 − π )u ′ (c2 (c1 ))c2′ (c1 ) = 0

−πu ′ (c1 )
=⇒ c2′ (c1 ) = .
(1 − π )u ′ (c2 (c1 ))
▶ Slope on the 45-line determined only by ratio of probabilities,

c1 = c2 (c1 )
−π
=⇒ c2′ = .
(1 − π )

36 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ Can use to model demand for insurance

▶ For example think of a possible loss in a state 1 (e.g. beach
front house permanently destroyed)
▶ State 1 is destroyed, state 2 is not destroyed.
▶ Interest is in decision on whether or not to insure it, and if so
how much insurance to purchase.
▶ Total wealth is W .
▶ Loss in state 1 is L < W .
▶ Endowment point is (W − L, W ) .

37 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

Figure: endowment: no insurance

38 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ Insurance is known as ”Arrow security”: pays off $Y if state 1

occur for premium amount $γY up front with γ ∈ (0, 1)
▶ Premium is paid before state is realized, i.e., in both states but
insurance pays off only in state 1
▶ On net purchasing $Y of such insurance lowers consumption by
$γY in state 2, and on net raises it by $(1 − γ)Y dollars in
state 1
▶ Note that low values of γ correspond to cheaper insurance.

39 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ For example suppose γ = 0.1 , a ten cent up front premium

purchases net 90c in state 1.
▶ Hence budget line reflects that 10c fall in state 2 produces 90c
in state 1
▶ Slope is - 19 .
▶ Generally it is - 1−γ γ .

40 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ Optimal choice where IC and price line are tangent

▶ Mathematically, at tangency

πu ′ (c1 ) γ
c2′ (c1 ) = − = −
( 1 − π ) u ′ ( c2 ) 1−γ

41 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

Figure: buying insurance: γ premium

42 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ Insurance is actuarially fair if expected payout equals up front

payment.
▶ For $1 final payout, probability π , pay up front premium of γ .
▶ Hence fair insurance requires γ = π .
▶ Actuarially unfair if up front payment above expected payout ,
γ>π.

43 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

Figure: red: actuarially unfair, blue: actuarially fair 44 / 128

 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ Availability of insurance gives rise to a feasible set of contingent

consumption levels
▶ Interest is in optimal purchase
▶ On 45-degree line called full insurance.
▶ Branch to right of 45-degree corresponds to over-insurance.
▶ Branch between endowment and 45-degree to partial insurance.
▶ Branch to left of endowment corresponds to selling insurance.
▶ Will see that wacky branches are not used.

45 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

Figure: regions of insurance on the budget line

46 / 128
 3.3 Insurance 3.3.1 Optimal Choice of Insurance

3.3.1 Optimal Choice of Insurance

▶ Using the tangency condition, for actuarially fair insurance
π γ
− =− ,
(1 − π ) 1−γ
u ′ ( c1 )
⇒ ′ =1
u ( c2 )
⇒ c1 = c2

i.e., full insurance

▶ Also geometrically implies partial insurance if actuarially unfair.
π γ
π < γ =⇒ − >−
(1 − π ) 1−γ

47 / 128
3.3 Insurance 3.3.2: Example

3.3.2: Example

48 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

▶ Example of optimal purchase of actuarially unfair insurance

▶ Total wealth is $24,000.
▶ Loss in state 1 is $18,000. Probability is π = 0.5 .
▶ In state 2 lose nothing have full $24,000. Probability is π = 0.5 .
▶ Endowment point is

(W − L, W ) = (6, 000, 24, 000).

49 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

▶ Insurance premium: pay $0.75 up front per dollar of payout in

state 1.
▶ Since
0.5 = π < γ = 0.75
this is actuarially unfair.
▶ Set up general optimization to determine how much to buy
▶ Need to be endowed with enough in both states to purchase
insurance.
▶ Else Demand Zero

50 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

▶ Let Y be dollars of insurance paid out in state 1, 0 ≤ Y .

▶ Given premiums, pay in 0.75Y to receive this.
▶ If Y is purchased, consume $[6, 000 + 0.25Y ] in state 1,
[24, 000 − 0.75Y ] in state 2.
▶ Resulting EU is therefore,

V (Y ) = 0.5u [6, 000 + 0.25Y ] + 0.5u [24, 000 − 0.75Y ]

▶ Note that,
1 3
V ′ (Y ) = u ′ [6, 000 + 0.25Y ] − u ′ [24, 000 − 0.75Y ]
8 8

51 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

▶ Objective is to maximize this on 0 ≤ Y ≤ 32, 000 : ignore upper

bound since optimum is partial insurance (π < γ).
▶ Note that concave function so the condition for interior is that
V ′ (Ŷ ) = 0 , together with the corner conditions that 0 is
optimal if V ′ (0) ≤ 0 .

52 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example
▶ Most common risk averse utility function is log form,

u (c ) = ln c.

▶ Solve now for special case of log utility u (c ) = ln c so that

u ′ (c ) = c1 and look for an interior solution with V ′ (Ŷ ) = 0 .
1 3
  =  ;
8 6, 000 + .25Ŷ 8 24, 000 − 0.75Ŷ

or, multiplying both sides by 8 times the product of the

denominators,

24, 000 − 0.75Ŷ = 18, 000 + 0.75Ŷ =⇒ Ŷ = $4, 000.

53 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example
▶ Optimally consume

ĉ1 = 6, 000 + 0.25Ŷ = $7, 000

in state 1 and,

ĉ2 = 24, 000 − 0.75Ŷ = $21, 000

in state 2.
▶ Transaction: out of $24,000 wealth, pay $3,000 up front to
purchase $4,000 of insurance. This leaves $21,000. This is
wealth in state 2. In state 1 if loss of $18,000 happens left with
only $3,000 but also get insurance payout of $4,000 hence end
up with $7,000 to consume in state 1
54 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

Figure: optimal insurance

55 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

▶ Note log form of utility makes example Cobb-Douglas like

▶ Hence there is a Cobb-Douglas explanation based on prices and
expenditure shares

56 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

▶ Thought experiment about selling entire contingent endowment

on market in period 0.
▶ Wealth in state 1 three times as valuable as wealth in state 2
according to insurance contract (75 cents given up in state 2
nets 25 cents net gain in state 1.
▶ Hence period 0 value of endowment in terms of units of state 2
consumption (dollars) if could sell full endowment,

p1 (W − L) + p2 W = 3 ∗ 6, 000 + 24, 000 = $42, 000.

57 / 128
 3.3 Insurance 3.3.2: Example

3.3.2 Example

▶ With equal expenditure shares, spend equal amounts on both

forms of consumption,

p1 ĉ1 = p2 ĉ2 = $21, 000

▶ With p1 = $3; p2 = $1 get,

ĉ1 = $7, 000; ĉ2 = $21, 000,

as we found.
▶ Other example in homework

58 / 128
3.4 Portfolio Allocation

3.4 Portfolio Allocation

59 / 128
3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

60 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ Another crucial application: portfolio allocation

▶ Total wealth W
▶ Allocate between two assets, one risky, one riskless
▶ Risky asset: gives a return of y% in good state and -z% in the
bad state
▶ Riskfree asset: gives return r% irrespective of states

61 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ Need to solve how to allocate W across two assets to maximize

utility
▶ Suppose x proportion is allocated to risky asset and 1 − x to
risk-free asset
▶ Return from risk-free asset (1 − x )W (1 + r )
▶ Return from risky asset xW (1 + y ) is good state and
xW (1 − z ) in bad state

62 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ π: probability of good state

▶ Total asset in good state: xW (1 + y ) + (1 − x )W (1 + r )
▶ Total asset in bad state: xW (1 − z ) + (1 − x )W (1 + r )
▶ Expected utility:

πu (xW (1 + y ) + (1 − x )W (1 + r ))
+ (1 − π )u (xW (1 − z ) + (1 − x )W (1 + r ))

▶ Note that, you can’t write riskless asset separately

63 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ Choose x ∈ [0, 1] to maximize EU

▶ Solve

max πu (xW (1 + y ) + (1 − x )W (1 + r ))
x
+ (1 − π )u (xW (1 − z ) + (1 − x )W (1 + r ))
▶ FOC

πu ′ (xW (1 + y ) + (1 − x )W (1 + r ))(W (1 + y ) − W (1 + r ))
+(1 − π )u ′ (xW (1 − z ) + (1 − x )W (1 + r ))(W (1 − z ) − W (1 + r ))
=0

64 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ Simplifying we get,

πu ′ (xW (1 + y ) + (1 − x )W (1 + r ))W (y − r )
= (1 − π )u ′ (xW (1 − z ) + (1 − x )W (1 + r ))W (z + r )

▶ Solving this will give optimal x

65 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ Example 1: log vNM utility function

▶ y = 0.15, z = 0.1, r = 0.05, π = 0.5, W = 100
▶ FOC becomes
10 15
=
115x̂ + 105(1 − x̂ ) 90x̂ + 105(1 − x̂ )
2(105 − 15x̂ ) = 3(10x̂ + 105)
30x̂ = −105
x̂ =0

▶ Since x̂ ∈ [0, 1]

66 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ Example 2: log vNM utility function

▶ y = 0.2, z = 0.1, r = 0.05, π = 0.8, W = 100
▶ FOC becomes
15 ∗ 0.8 15 ∗ 0.2
=
120x̂ + 105(1 − x̂ ) 90x̂ + 105(1 − x̂ )
4(105 − 15x̂ ) = (15x̂ + 105)
75x̂ = 375
x̂ =1

▶ Since x̂ ∈ [0, 1]

67 / 128
 3.4 Portfolio Allocation 3.3.6: Optimal Portfolio Choice

3.3.6: Optimal Portfolio Choice

▶ Example 3: log vNM utility function

▶ y = 0.2, z = 0.04, r = 0.05, π = 0.4, W = 100
▶ FOC becomes
15 ∗ 0.4 9 ∗ 0.6
=
120x̂ + 105(1 − x̂ ) 96x̂ + 105(1 − x̂ )
6(105 − 9x̂ ) = 5.4(105 + 15x̂ )
135x̂ = 0.6 ∗ 105 = 63
x̂ = 0.467

68 / 128
 3.5* Understanding Expected Utility

3.5* Understanding Expected Utility

69 / 128
3.5* Understanding Expected Utility 3.5.1: Construction of the EU Function

3.5.1: Construction of the EU Function

70 / 128
 3.5* Understanding Expected Utility 3.5.1: Construction of the EU Function

3.5.1 Construction of the EU Function

▶ A constructive procedure identifies an EU function for any finite

list of prizes.
▶ For simplicity suppose pure prizes are not indifferent and order
from best to worst.
▶ The utility of the best prize z1 is set to one, u (z1 ) = 1 , that of
the worst u (zN ) = 0 is set to zero.
▶ Recall in two good case that strict preference for one offering
higher chance of better prize.

71 / 128
 3.5* Understanding Expected Utility 3.5.1: Construction of the EU Function

3.5.1 Construction of the EU Function

▶ Now demonstrate EU theorem for three non-indifferent prizes

case
▶ Set u (z1 ) = 1 and u (z3 ) = 0.
▶ Find u (z2 ) = ū2 ∈ (0, 1) so that the second prize is indifferent
to the best with that probability and the worst with the
remaining probability,

z2 ∼ ū2 z1 + (1 − ū2 )z3 .

▶ Exists by A1
▶ Unique by strict monotonicity

72 / 128
 3.5* Understanding Expected Utility 3.5.1: Construction of the EU Function

3.5.1 Construction of the EU Function

▶ By substitution axiom, consider an arbitrary lottery (p1 , p2 , p3 ) ,
and substitute for prize 2 the indifferent lottery between best
and worst;
(p1 , p2 , p3 ) ∼ (p1 + p2 ū2 , 0, p3 + p2 (1 − ū2 )).
▶ Illustrate in tree diagram

Figure: Equivalent lotteries 73 / 128

 3.5* Understanding Expected Utility 3.5.1: Construction of the EU Function

3.5.1 Construction of the EU Function

▶ Now consider an alternative lottery (q1 , q2 , q3 ) , and again

substitute for prize 2 the indifferent lottery between best and
worst;

(q1 , q2 , q3 ) ∼ (q1 + q2 ū2 , 0, q3 + q2 (1 − ū2 )).

▶ Illustrate in tree diagram

74 / 128
 3.5* Understanding Expected Utility 3.5.1: Construction of the EU Function

3.5.1 Construction of the EU Function

▶ We know that among two prize lotteries the condition depends

only on probability of better prize,

(p1 + p2 ū2 , 0, p3 + p2 (1 − ū2 )) ≿ (q1 + q2 ū2 , 0, q3 + q2 (1 − ū2 ))

⇐⇒ p1 + p2 ū2 ≥ q1 + q2 ū2 ;

or,
3 3
Eu (p) = ∑ pi u (zi )) ≥ ∑ qi u (zi )) = Eu (q).
i =1 i =1
▶ Hence we can characterize preferences over the entire space in
an incredibly simple fashion, using only three numbers.

75 / 128
 3.5* Understanding Expected Utility 3.5.1: Construction of the EU Function

3.5.1 Construction of the EU Function

▶ Note after setting utility of worst strictly below best, rest implied
by indifference to corresponding worst-best lottery
▶ Meaning of utility numbers: a scale for evaluating differences in
utility is probability difference in best-worst lottery space
▶ Impacts cardinal-ordinal discussion

76 / 128
3.5* Understanding Expected Utility 3.5.2: Critiques of the EU Theorem

3.5.2: Critiques of the EU Theorem

77 / 128
 3.5* Understanding Expected Utility 3.5.2: Critiques of the EU Theorem

3.5.2: Critiques of the EU Theorem

▶ Behavioral economics built in part on critique of EU theory

▶ One problematic pattern for EU theorem:
▶ A1 = [($6000; 0.45), ($0; 0.55)] vs.
A2 = [($3000; 0.9), ($0; 0.1)] .
▶ B1 = [($6000; 0.001), ($0; 0.999)] vs.
B2 = [($3000; 0.002), ($0; 0.998)] .
▶ KT find that popular winners are A2 (86%) and B1 (73%)
▶ Violates the substitution axiom, since B1 = 4501 A1 + 449 [0; 1.0]
450
1 449
and B2 = 450 A2 + 450 [0; 1.0] .
▶ KT believe arises because of the ”possibility” effect.

78 / 128
 3.5* Understanding Expected Utility 3.5.2: Critiques of the EU Theorem

3.5.2: Critiques of the EU Theorem

▶ EU algebra: B1 ≺ B2 implies:

0.45u (6n ) + 0.55u (0) < 0.9u (3n ) + 0.1u (0).

▶ Subtraction and multiplication yields,

0.45u (6n ) + 0.45u (0) < 0.9u (3n );

u (6n ) + u (0) < 2u (3n )

▶ A1 ≻ A2 implies:

0.001u (6n ) + 0.999u (0) > 0.002u (3n ) + 0.998u (0);

u (6n ) + u (0) > 2u (3n ).

▶ The contradiction is clear.

79 / 128
 3.5* Understanding Expected Utility 3.5.2: Critiques of the EU Theorem

3.5.2: Critiques of the EU Theorem

▶ Another problematic pattern for EU theorem:

▶ C 1 = [($2500; 0.33), ($2300; 0.66), ($0; 0.01)] vs.
C 2 = [($2300; 1.0)] .
▶ D1 = [($2500; 0.33), ($0; 0.67)] vs.
D2 = [($2300; 0.34), ($0; 0.66)] .
▶ KT indicate popular winners are C2 (82%), D1 (83%).

80 / 128
 3.5* Understanding Expected Utility 3.5.2: Critiques of the EU Theorem

3.5.2: Critiques of the EU Theorem

▶ Yet if EU theorem valid, preference for C 2 over C 1 implies

preference for D2 over D1 :

C2 ≻ C 1 =⇒ u (2300) > 0.33u (2500) + 0.66u (2300) + 0.01u

=⇒ 0.33u (2300) > 0.33u (2500) + 0.01u (0)
=⇒ 0.33u (2300) + 0.66u (0) > 0.33u (2500) + 0.67u (0),
=⇒ D2 ≻ D1.

▶ KT ”certainty” effect.

81 / 128
 3.5* Understanding Expected Utility 3.5.2: Critiques of the EU Theorem

3.5.3: In Support of the EU Theorem

▶ Note that the critiques involve the claim that ”more matters
than money per se”
▶ Then they should be included as prizes!
▶ Models with ”psychological prizes” that include e.g. fear and an
avoidance decision
▶ EU theory with imperfect comprehension is the most important
model in an evolving approach to economics and psychology

82 / 128
 3.5* Understanding Expected Utility 3.5.2: Critiques of the EU Theorem

Unit 3: Risk and Time Preference

▶ Next two lectures on life-cycle framework

▶ Key to labor market and financial decisions
▶ 3.6 Basic Life Cycle Model
▶ 3.7 Uncertain Future Income and Precautionary Savings

83 / 128
3.6 Basic Life Cycle Model 3.6.1 Time Preference

3.6 Basic Life Cycle Model

3.6.1 Time Preference

84 / 128
 3.6 Basic Life Cycle Model 3.6.1 Time Preference

3.6.1: Time Preference

▶ Life Cycle EU model fantastically important.

▶ Consumption today vs consumption tomorrow
▶ Life cycle over T periods
▶ Consumption vector over life cycle (c1 , c2 , . . . cT )
▶ Utility function over consumption vectors ct

85 / 128
 3.6 Basic Life Cycle Model 3.6.1 Time Preference

3.6.1 Time Preference

▶ Standard form involves utility function separable across time

▶ Also future consumption is discounted relative to present with
discount factor β ∈ (0, 1)
▶ $1 tomorrow is as valuable as $β today

86 / 128
 3.6 Basic Life Cycle Model 3.6.1 Time Preference

3.6.1 Time Preference

▶ Standard assumption: Constant discount across time

▶ Called the exponential discounting model
▶ For example: the way DM discounts today vs tomorrow would
be the same today as 12 months from now
▶ Standard utility function under constant discounting and
separability,
T
u (c1 , c2 , .., cT ) = ∑ β t − 1 v ( ct ) .
t =1

87 / 128
 3.6 Basic Life Cycle Model 3.6.1 Time Preference

3.6.1 Time Preference

▶ We call v the period utility function

▶ Generally have v : R+ −→ R a strictly increasing concave
function (like in EU. )
▶ Notice that this makes it look a little bit like a risk averse EU
function.
▶ Treats time periods just like a particular state of world.

88 / 128
 3.6 Basic Life Cycle Model 3.6.1 Time Preference

3.6.1 Time Preference

▶ Different Interpretation: Concavity conveys preference for

equalizing consumption across periods.
▶ But not exact since the current period is more important than
future periods
▶ Relative importance of present-future defined by discount factor
β ∈ (0, 1)
▶ Present bias models: does not assume β constant across periods

89 / 128
 3.6 Basic Life Cycle Model 3.6.1 Time Preference

3.6.1 Time Preference

▶ Can illustrate in standard two good picture when T = 2.
▶ Change axes to consumption in two periods (c1 , c2 ).

Figure: utility function in two-period case

90 / 128
3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2 Intertemporal Budget Set

91 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ Endowment (m1 , m2 ) is income in each period

▶ Budget set depends on the access to credit market
▶ Draw budget set four ways:
▶ Saving but not borrowing: no interest (no credit market)
▶ Saving at interest rate r ( so that $1 → $(1 + r )) , no
borrowing
▶ Saving at interest r , borrowing at higher rate
▶ Borrowing and lending at rate r (called perfect capital markets)

92 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ No credit market implies no borrowing

▶ But savings is possible (under the bed savings technology)
▶ For example: suppose (m1 , m2 ) = ($50, $50)
▶ Budget set would be:

c1 ≤ 50; c2 ≤ (50 − c1 ) + 50

▶ Left of endowment: slope −1

▶ vertical line at endowment

93 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

Figure: no credit market

94 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ No borrowing constraint: common assumption in macro

▶ Saving possible for an interest rate r ; i.e., every dollar spent
generates 1 + r tomorrow
▶ Budget set becomes:

c1 ≤ 50; c2 = (1 + r )(50 − c1 ) + 50

▶ For example if r = 0.1 (10% interest rate)

c1 ≤ 50; c2 ≤ 105 − 1.1c1

95 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

Figure: no borrowing but saving at r

96 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ Imperfect credit market: borrow at a higher rate r ′ , saving at a

lower rate r (r < r ′ )
▶ Most realistic assumption; especially for poor agents
▶ The budget constraint:
(
(1 + r )(50 − c1 ) + 50 for c1 ≤ 50
c2 =
(1 + r ′ )(50 − c1 ) + 50 for c1 > 50

97 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

Figure: borrow at r ′ , save at r ; r < r ′

98 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ Perfect capital market: borrow and save at the same rate

▶ Easiest assumption to work with (will use for rest of the analysis)
▶ The budget frontier:

c2 ≤ (1 + r )(50 − c1 ) + 50

▶ If r = 0.1 (10% interest rate):

c2 = 105 − 1.1c1

99 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

Figure: perfect capital market

100 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ With perfect capital markets, budget constraint is present value
of lifetime income
c2 m2
c1 + = m1 +
1+r 1+r
▶ Holds for any T
▶ Present value: value of future income/consumption from today’s
viewpoint
▶ Future value: value of present income/consumption from
tomorrow’s viewpoint
▶ Multiply both sides of budget set with (1 + r )

(1 + r )c1 + c2 = (1 + r )m1 + m2
▶ same budget constraint as future value of lifetime income
101 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ Consider now a simple case with perfect capital markets

▶ Study the impact of a change in the interest rate on the budget
set(c1 , c2 ).
▶ Change in interest has two effects:
▶ changes the rate of substitution of consumption between periods
▶ affects the total lifetime wealth
▶ income effect also called the wealth effect

102 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ In Slutsky decomposition note that the income effect depends on

whether a borrower or a lender
▶ For a borrower, a higher interest rate produces a negative
income effect (lower lifetime wealth)
▶ For a saver, a higher interest rate produces a positive income
effect (higher lifetime wealth)
▶ Pretty obvious age-related implications: retirees have income
reduced by with lower rate

103 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

Figure: price effect of increase in r ; left: borrower, right: saver

104 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

▶ Wealth effect can lead to change in asset position

▶ Increase in r : borrower can become saver if income effect
stronger, saver remains saver
▶ Decrease in r : saver can become borrower if income effect
stronger, borrower remains borrower

105 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

Figure: Slutsky decomposition: r increase, borrower remains borrower

106 / 128
 3.6 Basic Life Cycle Model 3.6.2 Intertemporal Budget Set

3.6.2: Intertemporal Budget Set

Figure: Slutsky decomposition: r increase, borrower becomes saver107 / 128

3.6 Basic Life Cycle Model 3.6.3 Life Cycle Optimal Choice

3.6.3 Life Cycle Optimal Choice

108 / 128
 3.6 Basic Life Cycle Model 3.6.3 Life Cycle Optimal Choice

3.6.3: Life Cycle Optimal Choice

▶ When T = 2 the optimal choice problem is as follows:

max U (c1 , c2 ) = v (c1 ) + βv (c2 )

c1 ,c2
c2 m2
s.t. c1 + = m1 +
(1 + r ) (1 + r )
▶ FOC same as before,

v ′ ( c1 )
MRS = = 1+r
βv ′ (c2 )
▶ FOC known as Euler Equation
▶ Solving Euler equation and budget constraints give (ĉ1 , ĉ2 )

109 / 128
 3.6 Basic Life Cycle Model 3.6.3 Life Cycle Optimal Choice

3.6.3: Life Cycle Optimal Choice

▶ For purposes of solution, easiest (as always) log utility,

v (c ) = ln c

▶ Take two period Cobb-Douglas example with weight 1 on c1 , β

on c2 .
▶ Maximize,
u (c1 , c2 ) = ln c1 + β ln c2 ,
subject to budget constraint
c2 m2
c1 + = m1 + ≡ W1 .
1+r 1+r
▶ Simple form for budget constraint reflects assumption of perfect
capital markets
110 / 128
 3.6 Basic Life Cycle Model 3.6.3 Life Cycle Optimal Choice

3.6.3: Life Cycle Optimal Choice

▶ Classical ES solution,
W1
p1 ĉ1 = ;
1+β
βW1
p2 ĉ2 = .
1+β

where  
1
(p1 , p2 ) = 1, .
1+r

111 / 128
 3.6 Basic Life Cycle Model 3.6.3 Life Cycle Optimal Choice

3.6.3: Life Cycle Optimal Choice

▶ Hence,
W1 (1 + r ) βW1
ĉ1 = ; ĉ2 =
1+β 1+β
▶ Interesting special case with
1
β= .
(1 + r )
▶ Plugging in we get
W
ĉ1 = ĉ2 =
1+β
▶ β = 1/(1 + r ) the DM and the market discounts the future
consumption similarly
▶ Homework deals with cases β ̸= (1+1 r ) .
112 / 128
3.7 Uncertain Future Income and Precautionary Savings

3.7 Uncertain Future Income and Precautionary

Savings

113 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ Model amendments?
▶ Allow for uncertainty about future income and spending needs
▶ Realistic full life cycle
▶ Allow for borrowing constraints
▶ Include institutions e.g. Social Security
▶ Allow for uncertain returns
▶ Allow for uncertain lifetime

114 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ Take up one simple amendment: uncertain future income

▶ To solve simplify by getting rid of discounting to β = 1
▶ Also zero interest on savings r = 0

115 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ First period income Y1 = Y .

▶ Second period income depends on realization of states
▶ In low income state which has probability pL ,

Y2 = Y − L

with L ∈ (0, Y ).
▶ In high income state the same period 2 as 1, Y2 = Y .

116 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings
▶ How to change optimization?
▶ Note that the constraint is that cannot borrow against uncertain
part of future income, maximum consumption is period 1 is
Y + Y − L = 2Y − L
▶ The life-cycle problem optimal choice can be solved as

max U (c1 , c2 ) = v (c1 ) + v (c2 );

c1 ∈[0,2Y −L]

2Y − L − c1 with probability pL ;
s.t. c2 =
2Y − c1 with probability 1 − pL .

with v (c ) strictly increasing and concave

117 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ Use EU feature to rewrite the optimal choice problem as,

max U (c1 ) ≡ v (c1 ) + pL v (2Y − L − c1 )

c1 ∈[0,2Y −L]
+(1 − pL )v (2Y − c1 );

118 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings
▶ To optimize look for interior solution where derivative of U (c1 )
equals zero,

U ′ (c1 ) ≡ v ′ (c1 ) − pL v ′ (2Y − L − c1 ) − (1 − pL )v ′ (2Y − c1 ) = 0;

▶ Optimality condition is,

v ′ (ĉ1 ) = pL v ′ (2Y − L − ĉ1 ) + (1 − pL )v ′ (2Y − ĉ1 ).

▶ This says that the optimal choice equalizes the expected value of
marginal utility

119 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings
▶ General idea is that marginal utility is very high at low levels of
consumption.
▶ Equalizing expected value of marginal utility would imply
consuming less in period 1 (compared to a no-uncertain future
income scenario)
▶ Additional saving due to uncertainty is called precautionary
saving
▶ However this effect depends very much on curvature of utility or
risk aversion.
▶ The more risk averse an individual is, the higher their
precautionary savings must be.
120 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ For example with log utility marginal utility is

1
v ′ (x ) = .
x

▶ Plot and note diminishing and strictly convex.

▶ What this means is that hold back on spending significantly due
to risk of low income in the second period

121 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7:Uncertain Future Income and Precautionary

Savings

▶ A simple numerical example.

▶ Take year 1 income Y1 = 1 (units say $100K)
▶ Let Y2 = 1 with probability 0.5 and potentially lose all income
and earn nothing in year 2 with probability 0.5. Expected income
$50K in period 2

122 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ First idea would be to take expected loss income of $50K off,

get $150K, divide between period 1 and 2: hence spend $75K in
period 1, save $125K
▶ If no loss in income spend $125K on consumption in period 2, if
loss in income then only get $25K in period 2.
▶ Plot MU picture to see that that expected MU of consumption
in period 2 higher than period 1 so save more.

123 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

Figure: same expected expenditure: not optimal

124 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings
▶ Now to get precise do the example with Y1 = Y2 = 1 with
probability 0.5 and Y2 = 0 with probability 1/2 with log utility.
▶ Return now to the example,
1 ′ 1
v ′ (ĉ1 ) = v (2Y − L − ĉ1 ) + v ′ (2Y − ĉ1 )
2 2
1 ′ 1 ′
= v (1 − ĉ1 ) + v (2 − ĉ1 ).
2 2
▶ With log utility: ĉ1 ∈ [0, 1] must satisfy:
v ′ (ĉ1 ) = 0.5v ′ (1 − ĉ1 ) + (1 − pL )v ′ (2 − ĉ1 );
1 1 1
= + ;
ĉ1 2(1 − ĉ1 ) 2(2 − ĉ1 )
125 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ Cross multiplication by 2ĉ1 (1 − ĉ1 )(2 − ĉ1 ) :

2(1 − ĉ1 )(2 − ĉ1 ) = ĉ1 (2 − ĉ1 ) + ĉ1 (1 − ĉ1 )

4 − 6ĉ1 + 2ĉ12 = 3ĉ1 − 2ĉ12 ;
4ĉ12 − 9ĉ1 + 4 = 0,

126 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

▶ Quadratic Formula:
√
9± (81 − 64)
ĉ1 = .
8
▶ Given ĉ1 ∈ [0, 1] the solution is
√
9 − 17 5
ĉ1 = < < 0.75.
8 8
▶ Higher risk aversion (typical of what is estimated in portfolio
data) can give rise to even higher precautionary savings.

127 / 128
 3.7 Uncertain Future Income and Precautionary Savings

3.7: Uncertain Future Income and Precautionary

Savings

Figure: same expected marginal utility: optimal choice 128 / 128