Expected Value, Expected Utility & the Allais and Ellsberg Paradoxes Psychology 466: Judgment & Decision Making Instructor: John Miyamoto 11/10/2015: Lecture 07-1 Note: This Powerpoint presentation may contain macros that I wrote to help me create the slides. The macros aren’t needed to view the slides. You can disable or delete the macros without any change to the presentation. Outline • Expected Value and Expected Utility: What's the Difference? • Allais Paradox ♦ Common consequence principle (a.k.a. Savage’s independence axiom or the sure-thing principle) ♦ Anticipated regret ♦ Nonlinear probability weighting • Ellsberg Paradox Psych 466, Miyamoto, Aut '15 Lecture probably ends here What Is the Expected Value of a Gamble? 2 Expected Value of a Gamble Psych 466, Miyamoto, Aut '15 Expected Value of a Gamble (Cont.): More General Version 3 Expected Value of a Gamble (cont.) Psych 466, Miyamoto, Aut '15 Would It Be Rational to be an Expected Value Maximizer? 4 Would It Be Rational to be an Expected Value Maximizer? Expected Value Maximizer: Someone who always prefers the gamble that has the higher expected value. • Discussions pro and con during the 18th and 19th century. Rich men wanting to know, which is the better bet? Psych 466, Miyamoto, Aut '15 Are You an Expected Value Maximizer? Concrete Example 5 Are You an Expected Value Maximizer” • I offer you a choice: Higher Risk Lower Risk • An expected value maximizer would choose Option 1. • EV( Option 1 ) > EV( Option 2 ) Psych 466, Miyamoto, Aut '15 Continuation of this Example 6 Are You an Expected Value Maximizer? (Cont.) • I offer you a choice: Option 1: 50% chance you win $10,010 50% chance you lose $10,000 Option 2: 50% chance you win $2 50% chance you lose $10 Intuitive argument in favor of Option 2: ♦ The pleasure of winning +$10,010 is smaller in absolute magnitude than the pain of losing -$10,000. ♦ The worst that can happen with Option 2 is the pain of losing -$10. ♦ What really matters is the subjective value of the outcomes, +$10,010, +$2, -$10, -$10,000. and not the objective monetary amounts. Psych 466, Miyamoto, Aut '15 St. Petersburg Paradox - Introduction 7 St. Petersburg Paradox Psych 466, Miyamoto, Aut '15 Illustration of the St. Petersburg Game 8 St. Petersburg Paradox (cont.) Psych 466, Miyamoto, Aut '15 Expected Value of the St. Petersburg Game 9 Expected Value of St. Petersburg Game is Infinite! Psych 466, Miyamoto, Aut '15 EU of St. Petersburg Game is Infinite 10 Expected Value of St. Petersburg Game is Infinite! Psych 466, Miyamoto, Aut '15 Does It Feel Right that the St. Petersburg Game is Infinitely Valuable? 11 Expected Value of St. Petersburg Game is Infinite! Would you give your total wealth for the opportunity to play the St. Petersburg game? If you are an expected value maximizer, you should be eager to pay everything you own for the opportunity to play the St. Petersburg Game just once. Psych 466, Miyamoto, Aut '15 Bernoulli's Utility Hypothesis 12 Expected Value & Expected Utility • Nobody is an expected value maximizer. Nobody always prefers the gamble with the higher expected value. • Daniel Bernoulli (1738): People maximize the expected utility of their choices; not the expected value of their choices. • Utility of X = subjective value of possessing or experiencing X • Next 200 years: Economic theory attempts to get rid of the concept of subjective value. Psych 466, Miyamoto, Aut '15 Expected Utility Hypothesis: Simplified Mathematical Statement 13 Expected Utility Hypothesis (Simplified Version) Let U(X) be the utility of X and let EU(G) be the expected utility of a gamble G. Expected Utility Hypothesis: There exists a function U such that: (i) for every pair of gambles G1 and G2 , G1 preferred to G2 iff EU(G1) > EU(G2) (ii) If G = (X1, p; X2, 1-p) is any lottery (for money), then EU(G) = pU(X1) + (1 - p)U(X2) • The Expected Utility (EU) Hypothesis is the claim that a rational agent must satisfy (i) and (ii). Psych 466, Miyamoto, Aut '15 Example: Calculating the EU of Two Gambles 14 Example: Calculating the EU(Option 1 ) & EU(Option 2 ) • I offer you a choice: Assume these are the utilities: Option 1: 50% chance you win $10,010 50% chance you lose $10,000 Option 2: 50% chance you win $2 50% chance you lose $10 U($10,010) U($2) U(-$10) U(-$10,000) = = = = +8,000 +1.8 -2.5 -10,000 • Calculate the Expected Utility of Each Option: ♦ EU( Option 1 ) = (½ 8000) + ( ½ (-10,000) ) = -1,000 Utils ♦ EU( Option 2 ) = (½ 1.8) + ( ½ -2.5 ) ♦ EU( Option 1 ) < EU( Option 2 ). = -3.5 Utils If you are an EU maximizer, you will choose Option 2. Psych 466, Miyamoto, Aut '15 Rationality Does NOT Demand that We Be Expected Value Maximizers 15 Rationality Does Not Demand that We Be Expected Value Maximizers • An insurance company is (approximately) an EV maximizer. An individual person is not an EV maximizer. Why? • Suppose an insurance company sells 10,000 auto insurance policies for $500/year each. ♦ ♦ Insurance company knows that the expected value of each policy is -$420. An individual auto accident might cost $2,000 to $1,000,000, but they happen rarely. • Is it rational to buy auto insurance? • There are many examples where reasonable people are NOT EXPECTED VALUE MAXIMIZERS. Psych 466, Miyamoto, Aut '15 Transition: Risk Aversion Is Related to Shape of Utility Function 16 1944 - 1947: The Birth of Expected Utility Theory • Daniel Bernoulli (1738): People maximize the expected utility of their choices; not the expected value of their choices. • Utility of X = subjective value of possessing or experiencing X • Next 200 years: Economic theory attempts to get rid of the concept of subjective value. • 1944 - 1947: Mathematical work of von Neumann & Morgenstern leads to the discovery of expected utility theory. Psych 466, Miyamoto, Aut '15 Preference Axioms - What Are They? 17 Preference Axioms – What Are They? • Preference Axioms for EU Theory: A set of assumptions about preference behavior which, if satisfied, imply that a decision maker is an EU maximizer (conforms to EU theory). • Transitivity is an example of a preference axiom. • Sure-thing principle (common consequence assumption) is another example of a preference axiom. (To be explained next.) • Preference axioms can be construed as a normative claim: This is how a rational agent ought to behave. • Preference axioms can be construed as a descriptive claim: This is how people actually behave. Psych 466, Miyamoto, Aut '15 Allais Paradox 18 Allais Paradox Allais, M. (1953). Le comportement de l'homme rationnel devant le risque: Critique des postulats et axiomes de l'école Americaine. Econometrica, 21, 503-546. typical Choice 1: Option A: Receive 1 million for sure. Option B: Receive Receive Receive typical 2.5 million, 1 million, 0, Choice 2: Option A': Receive 1 million, Option B': Receive 2.5 million, 10% chance 89% chance 1% chance 11% chance, otherwise $0. 10% chance, otherwise $0. • (Write student responses on the board.) • Typical choices: Choose A from Choice 1 and choose B' from Choice 2. Psych 466, Miyamoto, Aut '15 Ellsberg Paradox 19 Tabular Representation of the Allais Choices Choice 1: Option A: Receive 1 million for sure. (); Option B: Receive 2.5 million, 10% chance, Receive 1 million, 89% chance, Receive 0 , 1% chance Choice 2: Option A': Receive 1 million, 11% chance, otherwise $0. Option B': Receive 2.5 million, 10% chance, otherwise $0. () Chance of Outcome Choice 1 Choice 2 Psych 466, Miyamoto, Aut '15 10% 89% 1% Option A $1 $1 $1 Option B $2.5 $1 $0 Option A' $1 $0 $1 Option B' $2.5 $0 $0 Same Slide without the Opaque Grey Rectangles 20 Allais Paradox Is Based on Common Consequences Choice 1: Option A: Receive 1 million for sure. (); Option B: Receive 2.5 million, 10% chance, Receive 1 million, 89% chance, Receive 0 , 1% chance Choice 2: Option A': Receive 1 million, 11% chance, otherwise $0. Option B': Receive 2.5 million, 10% chance, otherwise $0. () Chance of Outcome Choice 1 Choice 2 Psych 466, Miyamoto, Aut '15 10% 89% 1% Option A $1 $1 $1 Option B $2.5 $1 $0 Option A' $1 $0 $1 Option B' $2.5 $0 $0 Statement of the Common Consequence Principle 21 Common Consequences Principle (Other Names: Sure-Thing Principle, Savage’s Independence Axiom) Common Consequence Principle: If two options have the same consequence given some outcome, then you should ignore this consequence. ♦ Base your choice on the aspects of the options that differ. Chance of Outcome Choice 1 Choice 2 Typical Choice 10% 89% 1% Option A $1 $1 $1 Option B $2.5 $1 $0 Option A' $1 $0 $1 Option B' $2.5 $0 $0 Typical Choice Psych 466, Miyamoto, Aut '15 Allais Paradox Violates the Common Consequence Principle 22 Psychological Explanations for the Allais Paradox Choice 1: Option A: Receive 1 million for sure. Option B: Receive 2.5 million, 10% chance Receive 1 million, 89% chance Receive 0 , 1% chance Choice 2: Option A': Receive 1 million, 11% chance, otherwise $0. Option B': Receive 2.5 million, 10% chance, otherwise $0. ----------------------------------------------------------------------• Class: Propose psychological explanations for the Allais Paradox. Psych 466, Miyamoto, Aut '15 Anticipated Regret – Explanation for Allais Paradox 23 Explaining the Allais Paradox in terms of Anticipated Regret Choice 1: Option A: Receive 1 million for sure. Option B: Receive 2.5 million, 10% chance Receive 1 million, 89% chance Receive 0 , 1% chance Choice 2: Option A': Receive 1 million, 11% chance, otherwise $0. Option B': Receive 2.5 million, 10% chance, otherwise $0. -----------------------------------------------------------------------• If you choose option B in choice 1 and get $0, you will feel intense regret. Choosing option A avoids the possibility of regret. • If you choose option B' in choice 2 and get $0, you will not feel regret for your decision because you could have gotten $0 with option A' as well. Psych 466, Miyamoto, Aut '15 What Circumstances Cause Feelings of Regret? 24 Comment: Decision-Related Emotion Negative Emotion Positive Emotion Disappointment Relief (receive bad outcome when you hoped for a good outcome) (receive good outcome when you feared a bad outcome) Regret Self-Congratulation (?) (receive a bad outcome when a different choice would have produced a much better outcome) (receive a good outcome when a different choice would have produced a much worse outcome) Psych 466, Miyamoto, Aut '15 Allais Paradox & Nonlinear Perception of Probability 25 Chance of Outcome Choice 1 Choice 2 10% 89% 1% Option A $1 $1 $1 Option B $2.5 $1 $0 Option A' $1 $0 $1 Option B' $2.5 $0 $0 • Hypothesis: Choices 1 and 2 differ in terms of anticipated regret. o o potential regret no potential regret Why Do People Have Allais-Type Preferences? Regret – comparison between what you have experienced and what you would have experienced if you made a different choice. Anticipated Regret – anticipating that a choice will create the possibility of regret. Psych 466, Miyamoto, Aut '15 Clean Version of This Slide 26 Chance of Outcome Choice 1 Choice 2 10% 89% 1% Option A $1 $1 $1 Option B $2.5 $1 $0 Option A' $1 $0 $1 Option B' $2.5 $0 $0 • Hypothesis: Choices 1 and 2 differ in terms of anticipated regret. o o potential regret no potential regret Why Do People Have Allais-Type Preferences? Regret – comparison between what you have experienced and what you would have experienced if you made a different choice. Anticipated Regret – anticipating that a choice will create the possibility of regret. Psych 466, Miyamoto, Aut '15 Clean Version of This Slide 27 Explaining the Allais Paradox in terms of Nonlinear Perception of Probability Choice 1: Option A: Receive 1 million for sure, 0% chance of receiving 0 dollars. Option B: Receive 2.5 million, 10% chance Receive 1 million, 89% chance Receive 0 , 1% chance Choice 2: Option A': Receive 1 million, 11% chance, otherwise $0. 89% chance of $0 Option B': Receive 2.5 million, 10% chance, otherwise $0. 90% chance of $0 ------------------------------------------------------------------------• In choice 1-A, the chance of $0 is 0%; in choice 1-B, it is 1%. In choice 2-A', the chance of $0 is 89%; in choice 2-B', it is 90%. • Psychologically, the difference between a 0% and 1% chance of $0 is greater than the difference between an 89% and 90% chance of $0. Psych 466, Miyamoto, Aut '15 Summary: Explanations of the Allais Paradox 28 Typical Preferences in the Allais Paradox Violate EU Theory Choice 1: Option A: Receive 1 million for sure. (); Option B: Receive 2.5 million, 10% chance, Receive 1 million, 89% chance, Receive 0 , 1% chance Choice 2: Option A': Receive 1 million, 11% chance, otherwise $0. Option B': Receive 2.5 million, 10% chance, otherwise $0. () Choice 1: EU( Option A ) = (0.10)U($1 mil ) + (0.89)U( $1 mil ) + (0.01)U($1 mil ) EU( Option B ) = (0.10)U($2.5 mil ) + (0.89)U( $1 mil ) + (0.01)U($0 mil ) Choice 2: EU( Option A ) = (0.10)U($1 mil ) + (0.89)U( $0 mil ) + (0.01)U($1 mil ) EU( Option B ) = (0.10)U($2.5 mil ) + (0.89)U( $0 mil ) + (0.01)U($0 mil ) Psych 466, Miyamoto, Aut '15 Time Permitting: Present the Ellsberg Paradox 29 Time Permitting: Discuss the Ellsberg Paradox • Who is Daniel Ellsberg? Psych 466, Miyamoto, Aut '15 Presentation of the Choices for the Ellsberg Paradox 30 Tuesday, November 10, 2015: The Lecture Ended Here Psych 466, Miyamoto, Aut '15 31