uc berkeley department of economics foundations of ...kariv/219a_risk.pdf · 4. hey, j. and c. omre...
TRANSCRIPT
![Page 1: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/1.jpg)
UC BerkeleyDepartment of Economics
Foundations of Psychology and Economics (219A)
Module I: Choice under Uncertainty
![Page 2: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/2.jpg)
Reading list
1. Camerer, C. (1995) “Individual Decision Making,” in Handbook of Exper-imental Economics. J. Kagel and A. Roth, eds. Princeton U. Press.
2. Starmer, C. (2000) “Developments in Non-Expected Utility Theory: TheHunt for a descriptive Theory of Choice under Risk,” Journal of EconomicLiterature, 38, pp. 332-382.
3. Harless, D. and C. Camerer (1994) “The Predictive Utility of GeneralizedExpected Utility Theories,” Econometrica, 62, pp. 1251-1289.
![Page 3: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/3.jpg)
4. Hey, J. and C. Omre (1994) “Investigating Generalizations of ExpectedUtility Theory Using Experimental Data,” Econometrica, 62, pp. 1291-1326.
5. Holt, C. and S. Laury (2002) “Risk Aversion and Incentive Effects,” Amer-ican Economic Review, 92, pp. 1644-1655.
6. Choi, S., R. Fisman, D. Gale, and S. Kariv (2007) “Consistency and Het-erogeneity of Individual Behavior under Uncertainty,” American EconomicReview, 97, pp. 1921-1938.
7. Halevy, Y. (2007) “Ellsberg Revisited: An Experimental Study”, Econo-metrica, 75, pp. 503-536.
![Page 4: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/4.jpg)
Background
• Decisions under uncertainty enter every realm of economic decision-making.
• Models of choice under uncertainty play a key role in every field of eco-nomics.
• Test the empirical validity of particular axioms or to compare the predictiveabilities of competing theories.
![Page 5: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/5.jpg)
Experiments à la Allais
• Each theory predicts indifference curves with distinctive shapes in the prob-ability triangle.
• By choosing alternatives that theories rank differently, each theory can betested against the others.
• The criterion typically used to evaluate a theory is the fraction of choicesit predicts correctly.
![Page 6: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/6.jpg)
Allais (1953) I
— Choose between the two gambles:
$25, 000.33%
A :=.66−→ $24, 000 B :=
1−→ $24, 000&.01
$0
![Page 7: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/7.jpg)
Allais (1953) II
— Choose between the two gambles:
$25, 000 $24, 000.33%
.34%
C := D :=&.067
&.066
$0 $0
![Page 8: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/8.jpg)
The Marschak-Machina probability triangle
1
HP
Increasing preference
LP 0
1
H, M, and L are three degenerate gambles with certain outcomes H>M>L
![Page 9: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/9.jpg)
A violation of Expected Utility Theory (EUT)
A
B
CD
1
HP
LP 0
1
EUT requires that indifference lines are parallel so one must choose either A and C, or B and D.
![Page 10: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/10.jpg)
Contributions
Results have generated the most impressive dialogue between observationand theorizing:
— Violations of EUT raise criticisms about the status of the Savage axiomsas the touchstone of rationality.
— These criticisms have generated the development of various alternativesto EUT, such as Prospect Theory.
![Page 11: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/11.jpg)
Results have generated the most impressive dialogue between observationand theorizing (Camerer, 1995):
— ...EU violations are much smaller (though still statistically significant)when subjects choose between gambles that all lie inside the triangle...
— ...due to nonlinear weighting of the probabilities near zero (as the rankdependent weighting theories and prospect theory predict)...
— ...the only theories that can explain the evidence of mixed fanning,violation of betweeness, and approximate EU maximization inside thetriangle...
![Page 12: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/12.jpg)
Rank-dependent expected utility (Quiggin, 1982)
Index consequences (x1, ..., xn) such that x1 is the worst and xn is thebest.
The weights for i = 1, ..., n− 1 are give by:
ωi = π(pi + · · ·+ pn)− π(pi+1 + · · ·+ pn)
and ωn = π(pn).
The predictions of the model depend crucially on the form of π(·). Onepossibility is an (inverted) S-shaped probability weighting function.
![Page 13: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/13.jpg)
Limitations
Choice scenarios narrowly tailored to reveal anomalies limits the usefulnessof data for other purposes:
— Subjects face extreme rather than typical decision problems designedto encourage violations of specific axioms.
— Small data sets force experimenters to pool data and to ignore individ-ual heterogeneity.
![Page 14: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/14.jpg)
Research questions
Consistency
— Is behavior under uncertainty consistent with the utility maximizationmodel?
Structure
— Is behavior consistent with a utility function with some special struc-tural properties?
![Page 15: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/15.jpg)
Recoverability
— Can the underlying utility function be recovered from observed choices?
Heuristics
— Can heuristic procedures be identified when they occur?
![Page 16: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/16.jpg)
A new experimental design
An experimental design that has a couple of fundamental innovations overprevious work:
— A selection of a bundle of contingent commodities from a budget set(a portfolio choice problem).
— A graphical experimental interface that allows for the collection of arich individual-level data set.
![Page 17: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/17.jpg)
The experimental computer program dialog windows
![Page 18: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/18.jpg)
![Page 19: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/19.jpg)
• The choice of a portfolio subject to a budget constraint provides moreinformation about preferences than a binary choice.
• A large menu of decision problems that are representative, in the statisticalsense and in the economic sense.
• A rich dataset that provides the opportunity to interpret the behavior atthe level of the individual subject.
![Page 20: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/20.jpg)
Rationality
Let {(pi, xi)}50i=1 be some observed individual data (pi denotes the i-thobservation of the price vector and xi denotes the associated portfolio).
A utility function u(x) rationalizes the observed behavior if it achieves themaximum on the budget set at the chosen portfolio
u(xi) ≥ u(x) for all x s.t. pi · xi ≥ pi · x.
![Page 21: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/21.jpg)
Revealed preference
A portfolio xi is directly revealed preferred to a portfolio xj if pi · xi ≥pi · xj, and xi is strictly directly revealed preferred to xj if the inequalityis strict.
The relation indirectly revealed preferred is the transitive closure of thedirectly revealed preferred relation.
![Page 22: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/22.jpg)
Generalized Axiom of Revealed Preference (GARP) If xi is indirectlyrevealed preferred to xj, then xj is not strictly directly revealed preferred(i.e. pj · xj ≤ pj · xi) to xi.
GARP is tied to utility representation through a theorem, which was firstproved by Afriat (1967).
![Page 23: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/23.jpg)
Afriat’s Theorem The following conditions are equivalent:
— The data satisfy GARP.
— There exists a non-satiated utility function that rationalizes the data.
— There exists a concave, monotonic, continuous, non-satiated utilityfunction that rationalizes the data.
![Page 24: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/24.jpg)
Goodness-of-fit
• Verifying GARP is conceptually straightforward but it can be difficult inpractice.
• Since GARP offers an exact test, it is necessary to measure the extent ofGARP violations.
• Measures of GARP violations based on three indices: Afriat (1972), Varian(1991), and Houtman and Maks (1985).
![Page 25: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/25.jpg)
Afriat’s critical cost efficiency index (CCEI) The amount by whicheach budget constraint must be relaxed in order to remove all violationsof GARP.
The CCEI is bounded between zero and one. The closer it is to one, thesmaller the perturbation required to remove all violations and thus thecloser the data are to satisfying GARP.
![Page 26: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/26.jpg)
The construction of the CCEI for a simple violation of GARP
2x
1x
D
C
B A
x
y
The agent is ‘wasting' as much as A/B<C/D of his income by making inefficient choices.
![Page 27: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/27.jpg)
A benchmark level of consistency
A random sample of hypothetical subjects who implement the power utilityfunction
u(x) =x1−ρ
1− ρ,
commonly employed in the empirical analysis of choice under uncertainty,with error.
The likelihood of error is assumed to be a decreasing function of the utilitycost of an error.
![Page 28: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/28.jpg)
More precisely, we assume an idiosyncratic preference shock that has alogistic distribution
Pr(x∗) =eγ·u(x
∗)Rx:p·x=1
eγ·u(x),
where the precision parameter γ reflects sensitivity to differences in utility.
If utility maximization is not the correct model, is our experiment suffi-ciently powerful to detect it?
![Page 29: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/29.jpg)
The distributions of GARP violations – ρ=1/2 and different γ
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
CCEI
γ=1/4
γ=1/2
γ=1
γ=5
γ=10
![Page 30: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/30.jpg)
Bronnars’ (1987) test (γ=0)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
CCEI
n=5
n=15n=50
n=10
n=20
![Page 31: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/31.jpg)
Recoverability
• GARP imposes on the data the complete set of conditions implied byutility-maximization.
• Revealed preference relations in the data thus contain the information thatis necessary for recovering preferences.
• Varian’s (1982) algorithm serves as a partial solution to this so-called re-coverability problem.
![Page 32: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/32.jpg)
Risk neutrality
![Page 33: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/33.jpg)
Infinite risk aversion
![Page 34: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/34.jpg)
Loss / disappointment aversion
![Page 35: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/35.jpg)
The distributions of GARP violations - Afriat (1972) CCEI
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
00.0
5 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Frac
tion
of s
ubje
cts
Actual Random
![Page 36: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/36.jpg)
The distributions of GARP violations - Varian (1991)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
00.0
5 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Frac
tion
of s
ubje
cts
Actual Random
![Page 37: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/37.jpg)
The distribution of GARP violations - Houtman and Maks (1985)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50
Frac
tion
of s
ubje
cts
Actual
![Page 38: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/38.jpg)
Granularity
• A measure of the size of the components, or descriptions of components,that make up a system (Wikipedia).
• There is no taxonomy that allows us to classify all subjects unambiguously.
• A review of the full data set reveals striking regularities within and markedheterogeneity across subjects.
![Page 39: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/39.jpg)
ID 304 (0 / 1.000 / 50)
![Page 40: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/40.jpg)
ID 303 (0 / 1.000 / 50)
![Page 41: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/41.jpg)
ID 309 (17 / 0.952 / 48)
![Page 42: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/42.jpg)
ID 317 (0 / 1.000/ 50)
![Page 43: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/43.jpg)
ID 205 (0 / 1.000 / 50)
![Page 44: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/44.jpg)
ID 307 (12 / 0.916 / 46)
![Page 45: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/45.jpg)
ID 216 (0 / 1.000 / 50)
![Page 46: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/46.jpg)
ID 207 (15 / 0.981 / 47)
![Page 47: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/47.jpg)
ID 327 (5 / 0.965 / 49)
![Page 48: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/48.jpg)
ID 213 (0 / 1.000 / 50)
![Page 49: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/49.jpg)
Risk aversion
A “low-tech” approach of estimating an individual-level power utility func-tion directly from the data:
u(x) =x1−ρ
(1− ρ).
ρ is the Arrow-Pratt measure of relative risk aversion. The aversion to riskincreases as ρ increases.
![Page 50: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/50.jpg)
This generates the following individual-level econometric specification foreach subject n:
log
Ãxi2nxi1n
!= αn + βn log
Ãpi1npi2n
!+ i
n
where in ∼ N(0, σ2n).
We generate estimates of ρ̂n = 1/β̂n which allows us to test for hetero-geneity of risk preferences.
![Page 51: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/51.jpg)
The distribution of the individual Arrow-Pratt measures (OLS)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
0.22
0.24
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0Fr
actio
n of
sub
ject
s
OLS
![Page 52: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/52.jpg)
Loss/disappointment aversion
The theory of Gul (1991) implies that the utility function over portfoliostakes the form
min {αu (x1) + u (x2) , u (x1) + αu (x2)} ,
where α ≥ 1 measures loss/disappointment aversion and u(·) is the utilityof consumption in each state.
If α > 1 there is a kink at the point where x1 = x2 and if α = 1 we havethe standard EUT representation.
![Page 53: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/53.jpg)
An illustration of the derived demand
2
1lnxx
2
1lnpp
0ln2
1 =xx
0ln2
1 =pp
![Page 54: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/54.jpg)
The indifference map of Gul (1991)
1
HP
LP
1
![Page 55: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/55.jpg)
Scatterplot of the estimated CRRA parameters
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8
4.0
1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0α
ρ
![Page 56: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/56.jpg)
The risk premium r(1) for different values of α and ρ
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9h
r(h)
α=1, ρ =0
α=1, ρ =0.5
α=1.5, ρ =0.5
α=1.5, ρ =1.5
![Page 57: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/57.jpg)
Scatterplot of the risk measures – power and loss/disappointment
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8
4.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0r (1)
ρ
![Page 58: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/58.jpg)
Ambiguity
• The distinction between settings with risk and ambiguity dates back to atleast the work of Knight (1921).
• Ellsberg (1961) countered the reduction of subjective uncertainty to riskwith several thought experiments.
• A large theoretical literature (axioms over preferences) has developed mod-els to accommodate this behavior.
• But what matters most is the implications of the models for choice behav-ior.
![Page 59: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/59.jpg)
Ellsberg (1961)
An urn contains 300 marbles; 100 of the marbles are red, and 200 aresome mixture of blue and green. We will reach into this urn and select amarble at random:
— You receive $25, 000 if the marble selected is of a specified color.Would you rather the color be red or blue?
— You receive $25, 000 if the marble selected is not of a specified color.Would you rather the color be red or blue?
![Page 60: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/60.jpg)
Consider the following two-color Ellsberg-type urns (Halevy, 2007):
I. 5 red balls and 5 black balls
II. an unknown number of red and black
III. a bag containing 11 tickets with the numbers 0-10; the number writtenon the drawn ticket determines the number of red balls
IV. a bag containing 2 tickets with the numbers 0 and 10; the numberwritten on the drawn ticket determines the number of red balls.
![Page 61: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/61.jpg)
• A clever experiment to verify the connection between the reduction ofobjective compound lotteries and attitudes to ambiguity.
• Four different urns are used to elicit choices in the presence of risk, ambi-guity, and two degrees of compound uncertainty.
• Different models generate different predictions about how the reservationvalues (BDM) for these four urns will be ordered.
• The experiment can therefore classify each subject according to whichmodel predicts his ordering of reservation values.
![Page 62: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/62.jpg)
• Now, consider three states of nature and corresponding Arrow security(pays one dollar in one state and nothing in the other states).
• One state has an objectively known probability, whereas the probabilitiesof the other states are ambiguous.
• The presence of ambiguity could cause not just a departure from EU, buta more fundamental departure from rationality.
• Our analysis suggests otherwise — choices under ambiguity are at least asrationalizable as choices under risk.
![Page 63: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/63.jpg)
The distributions of CCEI scores
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
CCEI
Frac
tion
of s
ubje
cts
Risk Ambiguity Random
![Page 64: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/64.jpg)
Structure, recoverability and extrapolation
The conventional parametric approach:
— Choose a parametric form for the underlying utility function and fit theassociated demand function to the data.
— Test to see if they conform to the special restrictions imposed by hy-potheses concerning functional structure.
— Construct an estimate of the underlying utility function and forecastdemand behavior in new situations.
![Page 65: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/65.jpg)
Recursive Expected Utility (REU)
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the REU Model when alpha = 0.1
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the REU Model when alpha = 0.1
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the REU Model when alpha = 1
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the REU Model when alpha = 1
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
![Page 66: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/66.jpg)
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the REU Model when alpha = 1.75
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the REU Model when alpha = 1.75
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the REU Model when alpha = 2
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the REU Model when alpha = 2
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
![Page 67: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/67.jpg)
α-Maxmin Expected Utility (α-MEU)
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the Alpha-MEU Model when alpha = 0.55
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the Alpha-MEU Model when alpha = 0.55
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the Alpha-MEU Model when alpha = 0.6
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the Alpha-MEU Model when alpha = 0.6
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
![Page 68: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/68.jpg)
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the Alpha-MEU Model when alpha = 0.8
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the Alpha-MEU Model when alpha = 0.8
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p3)
x 1/(x 1+
x 3)
The Simulated x1/(x1+x3) and log(p1/p3) in the Alpha-MEU Model when alpha = 1
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
log(p1/p2)
x 1/(x 1+
x 2)
The Simulated x1/(x1+x2) and log(p1/p2) in the Alpha-MEU Model when alpha = 1
rho = 0.05
rho = 0.10
rho = 0.25rho = 0.5
rho = 1
rho = 2
![Page 69: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/69.jpg)
Individual-level data
![Page 70: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/70.jpg)
![Page 71: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/71.jpg)
![Page 72: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/72.jpg)
Estimation results
![Page 73: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/73.jpg)
![Page 74: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/74.jpg)
![Page 75: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/75.jpg)
Procedural rationality
• How subjects come to make decisions that are consistent with an underlyingpreference ordering?
• Boundedly rational individuals use heuristics in their attempt to maximizean underlying preference ordering.
— There is a distinction between true underlying preferences and revealedpreferences.
— Preferences have an EU representation, even though revealed prefer-ences appear to be non-EU.
![Page 76: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/76.jpg)
A type-mixture model (TMM)
A unified account of both procedural rationality and substantive rationality.
— Allow EU maximization to play the role of the underlying preferenceordering.
— Account for subjects’ underlying preferences and their choice of decisionrules.
![Page 77: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/77.jpg)
Ingredients
• The “true” underlying preferences are represented by a power utility func-tion.
• A discrete choice among the fixed set of prototypical heuristics, D, S andB(ω).
• The probability of choosing each particular heuristic is a function of thebudget set.
![Page 78: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/78.jpg)
• Subjects could make mistakes when trying to maximize EU by employingheuristic S.
• In contrast, when following heuristic D or B(ω) subjects’ hands do nottremble.
• A subject may prefer to choose heuristic B(ω) or D instead of the noisyversion of heuristic S.
![Page 79: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/79.jpg)
Specification
The underlying preferences of each subject are assumed to be representedby
u (x) =x1−ρ
(1− ρ)
(power utility function as long as consumption in each state meets thesecure level ω).
Let ϕ(p) be the portfolio which gives the subject the maximum (expected)utility achievable at given prices p.
![Page 80: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/80.jpg)
The ex ante expected payoff from attempting to maximize EU by employingheuristic S is given by
US(p) = E[πu (ϕ̃1(p)) + (1− π)u (ϕ̃2(p))]
ϕ̃(p) is a random portfolio s.t. p · ϕ̃(p) = 1 for every p = (p1, p2), andp1[ϕ̃1(p)− ϕ1(p)] = ε and i
n ∼ N(0, σ2n).
![Page 81: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/81.jpg)
When following heuristic D or B subjects’ hands do not tremble. Wetherefore write
UD (p) = u(1/(p1 + p2))
and
UB (p) = max{πu(0) + (1− π)u(1/p2), πu(1/p1) + (1− π)u(0)}
![Page 82: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/82.jpg)
Estimation
The probability of choosing heuristic k = D,S,B(ω) is given by a stan-dard logistic discrete choice model:
Pr(heuristic τ |p;β, ρ, σ) = eβUτPk=D,S,B
eβUk
where UD, US and UB is the payoff specification for heuristic D, S andB(ω), respectively.
![Page 83: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/83.jpg)
• The β̂ estimates are significantly positive, implying that the TMM hassome predictive power.
• Most subjects exhibit moderate to high levels of risk preferences aroundρ̂ = 0.8.
• There is a strong correlation between the estimated ρ̂ parameters from“low-tech” OLS and TMM estimations.
![Page 84: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/84.jpg)
The distribution of the individual Arrow-Pratt measures (TMM)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
0.22
0.24
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0Fr
actio
n of
sub
ject
s
TMM
![Page 85: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/85.jpg)
Goodness-of-fit
• Compare the choice probabilities predicted by the TMM and empiricalchoice probabilities.
• Nadaraya-Watson nonparametric estimator with a Gaussian kernel func-tion.
• The empirical data are supportive of the TMM model (fits best in thesymmetric treatment).
![Page 86: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/86.jpg)
Procedural rationality
• How subjects come to make decisions that are consistent with an underlyingpreference ordering?
• Boundedly rational individuals use heuristics in their attempt to maximizean underlying preference ordering.
— There is a distinction between true underlying preferences and revealedpreferences.
— Preferences have an EU representation, even though revealed prefer-ences appear to be non-EU.
![Page 87: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/87.jpg)
Archetypes and polytypes
• We identify a finite number of stylized behaviors, which collectively posea challenge to decision theory.
• We call these basic behaviors archetypes. We also find mixtures of archetypalbehaviors, which we call polytypes.
• The archetypes account for a large proportion of the data set and play arole in the behavior of most subjects.
• The combinations of types defy any of the standard models of risk aversion.
![Page 88: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/88.jpg)
Center
Vertex
![Page 89: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/89.jpg)
Centroid (budget shares)
Edge
![Page 90: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/90.jpg)
Bisector
Center and bisector
![Page 91: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/91.jpg)
Edge and bisector
Center, vertex, and edge
![Page 92: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/92.jpg)
Vertex and edge
Center and bisector
![Page 93: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/93.jpg)
The aggregate distribution of archetypes for different token confidence intervals
Center Vertex Centroid Edge Bisector All 0.1 0.005 0.003 0.000 0.019 0.083 0.110
0.25 0.061 0.004 0.002 0.083 0.187 0.337
0.5 0.093 0.011 0.007 0.139 0.215 0.466
1 0.134 0.032 0.019 0.165 0.252 0.602
2.5 0.185 0.064 0.049 0.186 0.285 0.769
![Page 94: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/94.jpg)
The distribution of archetypes, by subject (half token confidence interval)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Frac
tion
of d
ecis
ions
Center Vertex Centroid Edge Bisector
![Page 95: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/95.jpg)
The distribution of archetypes, by subject (one token confidence interval)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
5 90 22 53 47 46 77 68 63 58 32 12 45 49 98 66 38 91 18 8 95 41 81 88 69 94 75 23 67 7 62 92 57 40
Frac
tion
of d
ecis
ions
Center Vertex Centroid Edge Bisector
![Page 96: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/96.jpg)
A two- and three-asset experiment Token Shares in 3-asset experiment for Subject ID 3
TS2 = 1
TS1 = 1 TS3 = 1
0 10 20 30 40 50 60 70 80 90 1000
10
20
30
40
50
60
70
80
90
100
The relation of x1 and x2 in 2-asset experiment for ID 3
x1
x 2
![Page 97: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/97.jpg)
Token Shares in 3-asset experiment for Subject ID 47
TS2 = 1
TS1 = 1 TS3 = 1
0 10 20 30 40 50 60 70 80 90 1000
10
20
30
40
50
60
70
80
90
100
The relation of x1 and x2 in 2-asset experiment for ID 47
x1
x 2
![Page 98: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/98.jpg)
Token Shares in 3-asset experiment for Subject ID 25
TS2 = 1
TS1 = 1 TS3 = 1
0 10 20 30 40 50 60 70 80 90 1000
10
20
30
40
50
60
70
80
90
100
The relation of x1 and x2 in 2-asset experiment for ID 25
x1
x 2
![Page 99: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/99.jpg)
Token Shares in 3-asset experiment for Subject ID 61
TS2 = 1
TS1 = 1 TS3 = 1
0 10 20 30 40 50 60 70 80 90 1000
10
20
30
40
50
60
70
80
90
100
The relation of x1 and x2 in 2-asset experiment for ID 61
x1
x 2
![Page 100: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/100.jpg)
Token Shares in 3-asset experiment for Subject ID 65
TS2 = 1
TS1 = 1 TS3 = 1
0 10 20 30 40 50 60 70 80 90 1000
10
20
30
40
50
60
70
80
90
100
The relation of x1 and x2 in 2-asset experiment for ID 65
x1
x 2
![Page 101: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/101.jpg)
Type-mixture model (TMM)
• A subject chooses among the fixed set of types (heuristics), in order toapproximate the behavior that is optimal for his true underlying preferences.
• Consistent behavior requires subjects to choose among heuristics in a con-sistent manner as well as behaving consistently in applying a given heuristic.
• A TMM (CFGK, 2006) combines the distinctive types of behavior observedin the raw data in a coherent theory based on underlying preferences.
• It exhibits significant explanatory power and produces reasonable estimatesof risk aversion.
![Page 102: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/102.jpg)
Discussion
Suppose there are states of nature and associated Arrow securities andthat the agent’s behavior is represented by the decision problem
max u (x)s.t. x ∈ B (p) ∩A
where B (p) is the budget set and A is the set of portfolios correspondingto the various archetypes the agent uses to simplify his choice problem.
The only restriction we have to impose is that A is a pointed cone (closedunder multiplication by positive scalars), which is satisfied ifA is composedof any selection of archetypes except the Centroid.
![Page 103: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/103.jpg)
We can derive the following properties of the agent’s demand:
1. Let pk denotes the k-th observation of the price vector and
xk ∈ argmaxnu (x) : x ∈ B
³pk´∩A
odenotes the associated portfolio. Then the data
npk,xk
osatisfy
GARP.
2. There exists a utility function u∗ (x) such that for any price vector p,
x∗ ∈ argmax {u (x) : x ∈ B (p) ∩A}⇔
x∗ ∈ argmax {u∗ (x) : x ∈ B (p)} .
![Page 104: UC Berkeley Department of Economics Foundations of ...kariv/219A_risk.pdf · 4. Hey, J. and C. Omre (1994) “Investigating Generalizations of Expected Utility Theory Using Experimental](https://reader034.vdocuments.mx/reader034/viewer/2022043016/5f38f1699eb33d05e65a20b8/html5/thumbnails/104.jpg)
Takeaways
[1] Classical economics assumes that decisions are based on substantive ratio-nality, and has little to say about the procedures by which decisions arereached.
[2] Rather than focusing on the consistency of behavior with non-EUT theo-ries, we study the fine-grained details of individual behaviors in search ofclues to procedural rationality.
[3] The “switching” behavior that is evident in the data leads us to preferan “alternative” approach — one that emphasizes standard preferences andprocedural rationality.