Download presentation
Presentation is loading. Please wait.
Published byElaine Cain Modified over 9 years ago
1
Combining Bayesian Beliefs and Willingness to Bet to Analyze Attitudes towards Uncertainty by Peter P. Wakker, Econ. Dept., Erasmus Univ. Rotterdam (joint with Mohammed Abdellaoui & Aurélien Baillon) Decision and Uncertainty, Rotterdam, the Netherlands, April 11 '07 Topic: Uncertainty/Ambiguity. We make it operational; measuring, predicting, quantifying completely. Only gains today.
2
Good starting point for uncertainty: Risk. Allais: nonEU. Theories: 1.OPT (Kahneman &Tversky '79) 2.RDU (Quiggin '81) 3.Betweenness (Dekel, Chew '83, '86) 4.Quadratic utility (Chew, Epstein, Segal, '91) 5.Disappointment aversion (Gul '91) 6.Regret theory (Loomes, Sugden, Bell '82) 7.Skew-symmetric utility (Fishburn '87) 8.PT (Tversky & Kahneman '92) 2
3
All nonEU theories popular today: x y; xpy w(p)U(x) + ( 1–w(p) ) U(y); Relative to EU: one more graph … 3
4
4 inverse-S, (likelihood insensitivity) p w expected utility motivational cognitive pessimism extreme inverse-S ("fifty-fifty") prevailing finding pessimistic "fifty-fifty" Abdellaoui (2000); Bleichrodt & Pinto (2000); Gonzalez & Wu 1999; Tversky & Fox, 1997.
5
Models for risk came to a stop in 1990s. After Gilboa (1987) & Schmeidler (1989), Gilboa & Schmeidler (1989), and Tversky & Kahneman (1992): Now we turn to uncertainty/ambiguity (unknown probabilities). Still Allais nonEU. Theories: 1.Rank-dependent utility (CEU; Gilboa 1987; Schmeidler 1989); 2.Multiple priors (Wald 1950; Gilboa & Schmeidler 1989); 3.PT (Tversky & Kahneman 1992); 4.Endogeneous definitions of ambiguity …; 5.Smooth model (Klibanoff, Marinacci, Mukerji, 2005); 6.Variational model (Maccheroni, Marinacci, Rustichini, 2006). Economic models mostly normative. We: descriptive. 5
6
All popular static nonEU theories (except variational): x y; xEy W(E)U(x) + ( 1–W(E) ) U(y). (Ghirardato & Marinacci 2001; Luce 1991; Miyamoto 1988) EU: W is probability; for n states of nature, need n–1 assessments. General nonEU: need 2 n – 2 assessments. Pffff. No more graphs. Not "one more graph." Pffff. 6
7
Machina & Schmeidler (1992), probabilistic sophistication: x y; xEy w ( P(E) ) U(x) + ( 1–w ( P(E) ) ) U(y). This is doable. Relative to EU: one more graph! However, …. Ellsberg! 7
8
Common preferences between gambles for $100: (R k : $100) (R u : $100) (B k : $100) (B u : $100) > 8 Ellsberg paradox. Two urns with 20 balls. Ball drawn randomly from each. Events: R k : Ball from known urn is red. B k, R u, B u are similar. Known urn k 10 R 10 B 20 R&B in unknown proportion Unknown urn u ?20–? P(R k ) > P(R u ) P(B k ) > P(B k ) + 1 + 1 > < Under probabilistic sophistication with a (non)expected utility model:
9
Ellsberg: There cannot exist probabilities in any sense. Not "one more graph." 9 (Or so it seems?)
10
> Common preferences between gambles for $100: (R k : $100) (R u : $100) (B k : $100) (B u : $100) 20 R&B in unknown proportion Ellsberg paradox. Two urns with 20 balls. Ball drawn randomly from each. Events: R k : Ball from known urn is red. B k, R u, B u are similar. 10 R 10 B Known urn k Unknown urn u ?20–? P(R k ) > P(R u ) P(B k ) > P(B k ) ++ 1 1 > < Under probabilistic sophistication with a (non)expected utility model: 10 two models, depending on source reconsidered.
11
x y; xEy w S ( P(E) ) U(x) + ( 1– w S ( P(E) ) ) U(y). w S : source-dependent probability transformation. S: "Uniform" source. Ellsberg: w u (0.5) < w k (0.5) u k unknown known (Choice-based) probabilities can be maintained! Not one more graph. But a few more graphs. 11
12
Data: 12
13
13 0.125 0 0 Figure 8.3. Probability transformations for participant 2 Fig. a. Raw data and linear interpolation. 0.25 0.875 0.75 1 0.50 0.1250.875 0.25 0.50 0.751 Paris temperature; a = 0.78, b = 0.12 foreign temperature; a = 0.75, b = 0.55 risk: a = 0.42, b = 0.13 Within-person comparisons
14
14 participant 2; a = 0.78, b = 0.69 0 * Fig. a. Raw data and linear interpolation. * Figure 8.4. Probability transformations for Paris temperature 0.25 0.125 0.875 0.75 1 0.50 0.125 0.8750.2500.500.751 participant 48; a = 0.21, b = 0.25 Between-person comparisons
15
Example of predictions [Homebias; Within- Person Comparison; subject lives in Paris]. Consider investments. Foreign-option: favorable foreign temperature: $40000 unfavorable foreign temperature: $0 Paris-option: favorable Paris temperature: $40000 unfavorable Paris temperature: $0 Assume in both cases: favorable and unfavo- rable equally likely for subject 2; U(x) = x 0.88. Under Bayesian EU we’d know all now. NonEU: need some more graphs; we have them! 15
16
16 Paris temperature Foreign temperature decision weight expectation certainty equivalent uncertainty premium risk premium ambiguity premium 0.490.20 20000 177836424 135762217 5879 7697–3662 Within-person comparisons (to me big novelty of Ellsberg):
17
17 Subject 2, p = 0.125 decision weight expectation certainty equivalent uncertainty premium risk premium ambiguity premium 0.350.67 500035000 12133 159742732 5717 10257–3099 Subject 48, p = 0.125 Subject 2, p = 0.875 Subject 48, p = 0.875 500035000 0.080.52 2268 654 9663 –39 –4034 –7133 2078 9624 19026 25376 Between-person comparisons; Paris temperature
18
Conclusion: By (1) recognizing importance of uniform sources; (2) carrying out quantitative measurements of (a) probabilities (subjective), (b) utilities, (c) uncertainty attitudes (the graphs), all in empirically realistic and tractable manner, we make ambiguity completely operational at a quantitative level. 18
19
The end 19
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.