Chapter 1.

Decision Theory Under Uncertainty: Applications of the Expected Utility Theory

Advanced Microeconomics
3rd course (2nd semester) Degree in Economics and Finance Professor: Carmen Arguedas Departamento de Anlisis Econmico: T Econmica e H Econmica

CHAPTER 1 OUTLINE
1.1. Fair gambles and the expected utility hypothesis 1.2. The Von-Neumann-Morgenstern Theorem 1.3. Risk aversion 1.4. Applications of the expected utility theory

References: Nicholson (ch.7), Varian (ch.11), Pindyck & Rubinfield (ch.5).

1.1. Fair gambles and the expected utility hypothesis

In this chapter, we consider situations in which individuals (consumers, firms, the government, ...) make economic decisions under uncertainty. At the time individuals make the decision, the do not know the resulting outcome. Examples: Gambles, lotteries Insurance coverage Making transactions in the stock market Accepting/ rejecting labor offers Enrolling on business expansions (on new markets)

What are the questions we plan to answer in this chapter? How do we have to consider uncertainty to make economic decisions (such as consumption or investment decisions)?

For example, economic agents ask for loans to finance big purchases (housing, studies, cars, holidays) and plan to reimburse the money with (uncertain) future rents. How much risk are individuals willing to assume? For example, should individuals invest in risky assets or sure bonds? Should individuals accept a job offer or enrol in a degree or a master program? Etc.

What instruments can be used to answer the above questions?

Risk measures (to quantify the risk the of different alternatives)

Preferences for risk

Mechanisms to avoid or reduce risks

EXAMPLE 1 Suppose your income is 10.000 units and you buy a lottery ticket that offers a unique prize of 20.000 (out of 100 numbers). The cost of the ticket is 500 units. Analyze the following questions: 1. 2. 3. 4. 5. How much money you have under the two possible outcomes (winning / losing)? What is the probability (frequency) of each possible outcome? What is your expected money once you have bought the lottery ticket? What is the expected value of the lottery ticket? Do you think the price of the lottery ticket is fair?

6. Would you be willing to pay more or less than 500 to participate in this game?

EXAMPLE 2 Suppose your income is 100.000 euro, and you own a car valued 20.000 euro. There exists a 1% chance that your car is stolen. Analyze the following questions: 1. What is the expected value of your total wealth?

2. Supppose that you can buy insurance against robbery and you have two options: the first reimburses the entire value of the car in case of robbery costs and it costs 300 euro; the second reimburses only half of the value of the car in case of robbery and it costs 150 euro. Would you purchase any of these options?

3. Propose a fair insurance. Would you purchase it?

EXAMPLE 3 Are individuals always willing to accept fair games? St Petersburg paradox (due to Bernouilli)

Suppose you are offered the following game: You flip a coin as many times are neccesary until it comes up heads. If it comes up heads in the first try, you receive 2 euro; if it comes up heads in the second try, you receive 4 euro; if it comes up heads in the third try, you receive 8 euro; Etc. How much would you be willing to pay to play this game?

EXAMPLE 4

Suppose you receive two labor offers from two different companies. The first offer is a salary contingent on the companys profits: a) 2000 euro/ month if profits are large b) 1000 euro/month if profits are low The second offer is a fixed salary of 1510 euro/month. In this case, there exists a 1% chance that the company runs out of business. In this case, you would receive an unemployment benefit of 510 euro/month.

Analyze the two labor offers in risk terms. Which one would you choose? Why?

EXAMPLE 5

Assume instead the following two labor offers: In the first, the salary varies uniformely from 1000 to 2000 (for example, in increases of 100 euro, with equal probability). In the second, the salary varies uniformely between 1300 and 1700. Assume also there is no risk of unemployment in any of the two jobs. Analyze graphically the two labor offers in risk terms. Which one would you choose? Why?

MATHEMATICAL CONCEPTS

Random variable

x = {x1, x2, , xn}

or x [ x, x }

Probability density function

p = {p1, p2, , pn}, such that pi = 1 or f(x), such that f(x)dx = 1 E(x) = pixi or E(x) = xf(x)dx

Expected value of a random variable

Variance and standard deviation of a random variable V(x) = pi [xi- E(x)] 2 or E(x) = [x-E(x)] 2 f(x)dx

In example 5 above, both jobs pay the same salary on average. However, job 1 is riskier than job 2 since the dispersion of salaries (variance) in this case is greater.

What is a fair gamble?

A specified set of outcomes and associated probabilities with an expected value of zero.

Examples of fair gambles

Flipping a coin with a friend for a dollar Gaining 10 euros if a coin comes up heads nd losing 1 euro if it comes up tails with an entry fee of 4,5 euro.

Think of examples of unfair gambles Are individuals willing to take fair gambles?
St. Petersburg paradox

Bernouillis solution to the paradox:

Individuals do not care directly about money (dollar prizes of a gamble) BUT about the utility these dollars provide.

Assuming decreasing marginal utility, the St Petersburg gamble may converge to a FINITE expected utility value, even though its expected monetary value is infinite.

Thus, since the gamble provides a finite expected utility value, individuals would be willing to pay a FINITE amount of money to play it EXAMPLE 6 Compute the expeted utility value of the St Petersburg gamble assuming u(xi) = ln(xi).

1.2. The Von Neumann-Morgenstern Theorem

Rational individuals would make choices under uncertainty as if they had a utility function over money U(x) and maximized the expected value of U(x) (rather than the expected value of the payoff x itself)

Assume: GAMBLE A offers x1 with probability a and x2 with probability 1-a GAMBLE B offers x3 with probability b and x4 with probability 1-b

The VNM theorem says that a rational individual prefers gamble A to gamble B If and only if: a U(x1) + (1-a) U (x2) > b U(x3) + (1-b) U(x4)

Shortcomings (Varian, pages 192-193)

EXAMPLE 7 EXAMPLE 8

Allais paradox Ellsbergs paradox

1.3. Risk aversion

utility U(W) U(Wo) EU(A) GAMBLE A: gaining or losing h with equal probability (fair gamble)
Risk premium

Wo - h

CEA Wo

Wo + h

wealth

EXAMPLE 9 Compare gamble A with an alternative gamble B that offers the possibility of gaining or losing 2h with equal probability

EXAMPLE 10 Comment on example 7.2 from Nicholson, pages 196-197. (willingness to pay for insurance)

U U(W)

U U(W)

W Risk neutrality For example, U(W) = W

W Risk loving For example, U(W) = W, n > 1

How can we measure risk aversion?

Absolute risk aversion measure:

AR (W) =

U (W) U (W)

Relative risk aversion measure:

AR (W) = W

U (W) U (W)

1.4. Applications of the expected utility theory

Methods for reducing uncertainty and risk

Insurance Diversification Flexibility Information acquisition

Insurance

EXAMPLE 11

Consider an individual that has an initial wealth W and a property valued at L that is subject to a risk fire. In order to protect himself against the risk, the individual can buy an insurance policy from an insurance company. The company and the individual have the same prior as to the probability that the fire will occur (denoted as p). Assume the individual can insure the entire value of the property. The company demands a price for providing full coverage should the property burn down. This price is known as the premium, and we denote it by r. 1) What would be a fair premium? 2) Would the individual be willing to buy it? Assume the individual is risk averse and the utility function for wealth is U(W) = ln(W) 3) What is the maximum premium this individual would be willing to pay for full coverage againt the risk of fire?

Problems with insurance (for the insurance company)

Adverse selection: only the worst consumers (those who expect larger or more likely losses) may end up buying insurance

Moral hazard: having insurance may make consumers less willing to take steps to avoid losses

MORE ON CHAPTER 3. (ASYMMETRIC INFORMATION MODELS)

Diversification

Dont put all your eggs in one basket

Mainly for investment activities: Diversify your portfolio

EXAMPLE 12 An individual with wealth W can invest in two independent risky assets 1 and 2. Assume both assets have the same expected payoff and the same variance. Calculate the risk minimizing portfolio (or combination of assets).

Flexibility
Applicable to all-or-nothing decisions The decision maker can obtain some of the benefits of diversification by making flexible decisions. Payoff A2 Utility U (A2) U (A1)

A1 x State x x State x

EXAMPLE 13

Assume there are two alternatives A1 and A2 . The payoffs of these alternatives depend on the state of the world, which is a uniform variable on [0,1]. Assume A1 = 1-x and A2 = x. Assume also that the individual is risk neutral. a) What alternative should the individual choose if he or she did not know the state of the world at the time the decision has to be made? b) What is the maximum the individual would pay to have the option to choose the alternative once the state of the world is known?

Information acquisition

The value of information can be calculated as the difference between the utility of the choice made when the information is known and the expected utility of the choice made when the information is unknown.

It can be modelled similarly to the case of flexibility, studied above