Decision Making under Uncertainty and Subjective. Probabilities
|
|
- Wesley Lester
- 6 years ago
- Views:
Transcription
1 Decision Making under Uncertainty and Subjective Probabilities Edi Karni Johns Hopkins University September 7, 2005 Abstract This paper presents two axiomatic models of decision making under uncertainty that avoid the use of a state space The first is a subjective expected utility model with action-dependent subjective probabilities and effect-dependent preferences (the case of effect-independent preferences is obtained as a special instance) The second is a nonexpected utility model involving well-defined families of action-dependent subjective probabilities on effects and a utility representation that is not necessarily linear in these probabilities (a probabilistic sophistication version of this model, with actiondependent subjective probabilities is obtained as a special case) I am grateful to the National Science Foundation for financial support under grant SES
2 1 Introduction The notion of states of nature is a cornerstone of modern theories of decision-making under uncertainty Introduced by Savage (1954), a state of nature formalizes the idea of complete resolution of uncertainty that is, the assignment of a unique consequence to each conceivable course of action 1 For example, when betting on the outcome of a horse race, the states of nature correspond to the orders in which the horses may cross the finish line (these are the states of nature in Anscombe and Aumann [1963]) Theories of decision-making under uncertainty that invoke the notion of a state space (that is, the set of all states of nature) require that the states be so defined as to render the likely realization of alternative events (that is, subsets of the state space) independent of the decision maker s choice, and the valuation of the consequences independent of the states in which they may obtain Consequently, if the likely outcomes of a horse race may be affected by actions taken by a decision maker (for example, a jockey throwing a race) then the outcomes of the race no longer qualify as states of nature Similarly, taking out health insurance policy is betting on one s state of health The uncertainty involved is resolved once the true state of the insured s health becomes known However, the likely realization of alternative states of health is not independent of the life style (eg, diet and exercise regimen) adopted by the insured Moreover, in general, the insured s valuation of the indemnity is not independent of his state of health Hence the states of the decision maker s health do not 1 Formally, a state of nature is a function on the set of courses of action, or acts, to the set of consequences 2
3 qualify as states of nature If the framework of Savage is to be maintained, a different, more abstract, formulation of a state space is called for in which the outcomes of the horse race and the states of the decision maker s health become random variables However and here is the rub in many situations involving decision-making under uncertainty, the state space that meets these conditions is too abstract and/or too complex to contemplate Except in special circumstances, the state space does not correspond to an image of the world that decision makers invoke when making decisions under uncertainty 2 In Karni (2005) I advanced a new approach to modeling decision making under uncertainty This approach dispenses with the idea of a state space In this paper I pursue this approach using the Anscombe and Aumann (1963) device of roulette lotteries Taking advantage oftherichnessof thechoicespaceafforded by the availability of roulette lotteries, I develop axiomatic subjective expected utility models with unique family of action-dependent subjective probabilities and a family of effect-dependent utility functions that are simple and transparent Moreover, following Machina and Schmeidler (1992, 1995), I also develop an axiomatic model of decision-making under uncertainty in which action-dependent subjective probabilities are defined but the utility representation is not necessarily linear in these probabilities As a special case, I obtain a version of Machina and Schmeidler s probabilistically sophisticated choice model with action-dependent subjective probabilities A remarkable aspect of this new approach is that the family of subjective probabilities that figure in the representation is a proper representation of the decision makers introspec- 2 Additional examples, a more detailed discussion and references, are provided in Karni (2005) 3
4 tive beliefs, that is, his beliefs about the likely realization of events conditional on his actions that he uses to assess the alternative courses of action In the next section I describe the analytical framework The subjective expected utility theory is the subject matter of Section 3 Section 4 includes an exposition of the more general nonexpected utility models Concluding remarks appear in Section 5 and the proofs are given in Section 6 2 The Model 21 The Analytical Framework Let Θ be a finite set whose elements are effects In the examples cited earlier, effects are possible outcomes of a horse race or the states of a person s health Let A be a connected separable topological space set whose elements, referred to as actions, correspond to initiatives by which a decision maker believes that he might affect the likely realization of alternative effects (eg, effort level) Let Z (θ) be an arbitrary set of prizes that are feasible if the effect θ obtains To simplify the exposition assume that > Z (θ) 3 for every θ, and denote by (Z (θ)) the set of all probability distributions on Z (θ) Elements of (Z (θ)) are referred to as roulette lotteries, or simply lotteries Assume that, for each θ Θ, (Z (θ)) is endowed with the R n topology 4
5 Bets are effect-contingent lottery payoffs Formally, a bet, b, is a function on Θ such that b (θ) (Z (θ)) Denote by B the set of all bets (that is, B := Π θ Θ (Z (θ))) The choice set is the product set C := A B whose generic element, (a, b), is an action-bet pair Action-bet pairs represent conceivable alternatives among which decision makers may have to choose Assume that C is endowed with the product topology The set of consequences C consists of the prize-effect pairs, that is, C := {(z,θ) z Z (θ),θ Θ} Decision makers are characterized by preference relations, <, on C that have the usual interpretation 3 In other words, decision makers are supposed to be able to choose, or express preferences, among action-bet pairs presumably taking into account their beliefs regarding the influence that their choice of action may exert on the likely realization of alternative effects and, consequently, on the desirability of the corresponding bets The strict preference relation, Â, and the indifference relation,, are defined as usual For all b, b 0 B and α [0, 1], defined (αb +(1 α) b 0 )(θ) =αb (θ) +(1 α) b 0 (θ), for all θ Θ For p (Z (θ)) Iusethenotationb θ p to denote the bet that result from replacing the θ coordinate of b with the lottery p Denote by δ z the (degenerate) lottery that yields the prize z with probability one The prize z (θ; p) Z (θ) is said to be the certainty equivalent of p (Z (θ)) in θ, if 3 A preference relation, <, is a binary relation on C; (a, b) < (a 0,b 0 ) has the interpretation (a, b) is at least as desireable as (a 0,b 0 ) 5
6 a, b θ δ z(θ;α) (a, b θ p) I assume throughout that for all θ Θ and p (Z (θ)), the certainty equivalent of p in θ exists The preference relation < is said to be nondegenerate if there are b 0,b B such that b 0 Â b, otherwise the preference relation is degenerate Given <, an effect θ Θ is null given the action a if (a, b θ p) (a, b θ q) for all p, q (Z (θ)), and b B, otherwise it is nonnull given the action a Note that, given <, an effect may be null under some actions and nonnull under others Denote by Θ (a; <) the subset of effects that are nonnull given a 22 Continuous weak orders The first two axioms are part of all the models below These axioms are familiar and require no further explanation (A1) (Weak order) < is complete and transitive (A2) (Continuity) For all (a, b) C the sets {(a 0,b 0 ) (a 0,b 0 ) < (a, b)} and {(a 0,b 0 ) (a, b) < (a 0,b 0 )} are closed 6
7 23 Constant utility bets In general the welfare implications of the choice of actions are direct (for example, exertion of effort may be unpleasant), and indirect, through their perceived impact on the likely realization alternative effects Broadly speaking, a constant-utility bet is a bet whose effectcontingent payoffs entail compensating variations that render uniform the valuation of the ensuing effect-lottery pairs Thus, having accepted a constant utility bet, the decision maker is indifferent to which particular effect obtains Consequently, constant utility bets neutralize the indirect well-being implications of the actions The notion of constant-utility bet, introduced in Karni (2005a), is key to the theory presented here and to the claim that the subjective probabilities defined in this framework represent the decision makers beliefs The following additional axioms separate the direct well-being implications of actions and those of the constant-utility bets Let B be a convex subset of B and suppose that the restriction of < to A B satisfies the following well-known axioms: (A3) (Essentiality) There are a, a 0 A and b, b 0 B such that a, b  a 0, b and a, b  a, b0 (A4) (Coordinate independence) Foralla, a 0 A and b, b 0 B, a, b < a, b 0 if and only if a 0, b < a 0, b 0 and a, b < a 0, b if and only if a, b 0 < a 0, b 0 (A5) (Hexagon condition) For all a, a 0,a 00 A and b, b 0, b 00 B if a, b 0 a 0, b and a, b00 a 0, b 0 a 00, b then a 0, b 00 a 00, b 0 7
8 Theelementsof B have the interpretation of constant-utility bets The assumption that B is convex merits elaboration Denote by δ z := δ z(θ) θ Θ B the bet that assigns to θ the sure outcome z (θ) Suppose that δ z,δ z 0 B and (a, δ z 0) Â (a, δ z ) The the convexity of B implies that αδ z +(1 α) δ z 0 B, for all α [0, 1] Let z (θ, α) be the certainty equivalent of αδ z(θ) +(1 α) δ z 0 (θ), then, by the richness assumption below, δ z(α) := δ z(θ;α) θ Θ B for all α [0, 1] The convexity of B is responsible for the particular relation among the effect-dependent utility function in Theorem 1 below and, more importantly, the uniqueness of the action-dependent subjective probabilities I assume throughout that the constant utility bets are well-defined and for every given action each bet has an equivalent constant utility bet To formalize these assumption I introduce the following additional notations For every b B and a A, let I b; a = {b B (a, b) a, b } and let I (p; θ, b, a) ={q (Z (θ)) (a, b θ q) (a, b θ p)} Richness assumption: (a) b 0 a A I b; a if and only if b0 (θ) I b (θ);θ, b, a for all θ Θ and a A (b) For all (a, b) C there is b B such that a, b (a, b) To grasp the meaning of this assumption, note that distinct actions correspond to distinct beliefs regarding the likely realization of the different effects If A is not sufficiently rich, then given < there may exist b B and bets b a A I b; a (that is, a, b (a, b) for all a A) andyetb (θ) / a A I b (θ);θ, b, a for some θ Θ In this case, it is easy to conceive 8
9 of an action, ā, such that ā, b (ā, b) Hence if ā is added to A, then b/ a A {ā} I b; a The assumption states that the set of actions is sufficiently rich that there exist no additional actions that, if added to the set A, would reduce the set of bets that are indifferent to b conditional on every action in A The second part of the richness assumption requires that every bet has a certain-utility equivalent under all actions 3 Subjective Expected Utility Theory In this section I explore alternative subjective expected utility models 31 Effect-Dependent Preferences Conditional independence is the application of the independence axiom of expected utility theory to elements of the choice set that have the same action (A6) (Conditional Independence) Foralla A, b, b 0,b 00 B and α (0, 1], (a, b) < (a, b 0 ) if and only if (a, (αb +(1 α) b 00 )) < (a, (αb 0 +(1 α) b 00 )) Next I assume that, conditional of the effects, the risk attitudes displayed by the decision maker are independent of the actions This requires that the ranking of bets whose payoffs agree on all nonnull effects except one, be independent of the action (that is, the conditional preferences on lotteries given the effect are action-independent) Formally, 9
10 (A7) (Action independent risk attitudes) For all a, a 0 A, b B, θ Θ (a; <) Θ (a 0 ; <) and p, q (Z (θ)), (a, (b θ p)) < (a, (b θ q)) if and only if (a 0, (b θ p)) < (a 0, (b θ q)) 32 Subjective expected utility with effect-dependent preferences The first representation theorem below asserts that if the richness assumption holds then the axiomatic structure depicted by (A1) (A7) is necessary and sufficient for the existence of subjective expected utility representation with action-dependent probabilities and effectdependent risk attitudes Theorem 1 Let < be a nondegenerate preference relation on C and suppose that the richness assumption holds, then (a) The following conditions are equivalent: (ai) < satisfies (A1) (A7) (aii) There exists a family of probability measures {π ( ; a)} a A on Θ, continuous, jointly cardinal, real-valued functions u on C and v on A such that, for all (a, b), (a 0,b 0 ) C, (a, b) < (a 0,b 0 ) if and only if X π (θ; a) X π (θ; a 0 ) X u (z; θ) b 0 (z; θ)+v (a 0 ) (1) θ Θ u (z; θ) b (z; θ)+v (a) X z Z(θ) θ Θ 10 z Z(θ)
11 where u (λ (θ 0 ) z; θ 0 )=u (z; θ),λ(θ 0 ) > 0 for all θ 0 Θ {θ} and z Z (θ) (b) Thefamilyofprobabilitymeasures{π ( ; a) a A} on Θ is unique and π (θ; a) =0if and only if θ is null given a 33 Subjective expected utility with effect-independent preferences Consider next the case of effect-independent risk attitudes, that is, the case in which the ranking of lotteries is the same across effects This is analogous to the Anscombe and Aumann (1963) state-independence, or monotonicity, axiom Formally, (A8) (Effect independence) For all a A, b B, θ,θ 0 Θ (a; <), and p, q (Z (θ)) (Z (θ 0 )), (a, b θ p) < (a, b θ q) if and only if (a, b θ 0p) < (a, b θ 0q) The next theorem establishes that if the preference relation satisfies effect independence in addition to the other axioms, then the utility functions that figures in Theorem 1 are effect independent That is, not only do they display uniform attitudes towards risk, they also represent identical evaluation of the prizes Theorem 2 Let < be a preference relation on C and suppose that the richness assumption holds, then (a) The following conditions are equivalent: 11
12 (ai) < satisfies (A1) (A8) (aii) There exists a family of probability measures {π ( ; a) a A} on Θ, and jointly cardinal, continuous, real-valued functions u on θ Θ Z (θ) and v on A such that, for all (a, b), (a 0,b 0 ) C, (a, b) < (a 0,b 0 ) if and only if X π (θ; a) X π (θ; a 0 ) X u (z) b 0 (z; θ)+v (a 0 ) (2) θ Θ z Z(θ) u (z) b (z; θ)+v (a) X θ Θ z Z(θ) (b) Thefamilyofprobabilitymeasures{π ( ; a) a A} on Θ is unique and π (θ; a) =0if and only if θ is null given a Unlike in the traditional formulation of the subjective expected utility model, effectindependent preferences (or risk attitudes) imply effect-independent utility functions Hence in this case Ramsey s (1931) idea of defining the degree of belief in the truth of a proposition, or the likely realization of events, by the decision maker s betting behavior makes sense Otherwise, the decision maker s beliefs are confounded with his tastes in a way that renders the representation of these beliefs by subjective probabilities, representing the odds that he will accept, meaningless 4 4 For a detailed discussion of this issue see Karni (2005a) 12
13 4 Subjective Probabilities without Expected Utility In this section I weaken the conditional independence axiom to obtain a more general choicebased definition of subjective probabilities In particular, the model does not imply that the utility representation is linear in the probabilities 41 Motivation Machina and Schmeidler (1992, 1995) argue, convincingly, that a choice-theoretic definition of subjective probabilities does not require that the preference relation satisfy the axioms of subjective expected utility Invoking the analytical frameworks of Savage (1954) and Anscombe and Aumann (1963), respectively, they show that a decision maker may be probabilistically sophisticated, in the sense that his betting behavior implies the existence of unique subjective probabilities on the state space, his choice among acts is determined by his preferences among the induced distributions on the set of outcomes, and his preferences are representable by a utility function that is not necessarily linear in the probabilities However, as in expected utility theory, the definition of subjective probabilities in the theory of probabilistically sophisticated choice of Machina and Schmeidler is based in an implicit unverifiable assumption of state-independent outcome valuation My next objective is to develop, within the analytical framework of Section 2, a decision theory that yields a definition of action-dependent subjective probabilities on the effects 13
14 without requiring that the utility representation be linear in the probabilities Unlike the models of Machina and Schmeidler, the theory I propose does not require that the implicit valuation of the consequences be effect independent However, if constant bets are constant utility bets, a version of this model analogous to probabilistic sophistication is obtained Some modification of the lottery structure is introduced to simplify the exposition Let Θ = {θ 1,, θ n } and, for each θ Θ, let Z (θ) :=[zθ,z θ monetary prizes Let b = ³ δ z θ 1,,δ z θn and b = maximal and minimal elements of C, respectively 5 ³ δ z θ1,, δ z θ n ] be a real interval representing be bets in B that are the 42 Axioms To obtain the desired generalization, I replace conditional independence, (A6), that is responsible for the linear structure of the preference relation with two axioms that are analogous to Machina and Schmeidler s (1995) axioms 5 and 6 Given θ Θ and a lottery p (Z (θ)), denote by F p the cumulative distribution function corresponding to p Then, as usual, p is said to first-order stochastically dominate q if F q (z) F p (z) for all z Z (θ) This dominance relation is denoted p 1 q If, in addition, the inequality is strict for some z Z (θ), then the dominance relation is strict and is denoted p> 1 q 5 This assumption is analogous to the assumption of Machina and Schmeidler (1995) that the set of outcomes includes a best and worst outcome 14
15 The axiom of conditional monotonicity asserts that, given a A, b B, and θ Θ (a; <), the restriction of < to {(a, b θ p) p (Z (θ))} is strictly monotonic with respect to first-order stochastic dominance Formally, (A9) (Conditional monotonicity) For all a A, θ Θ, b B, and p, q (Z (θ)), if p 1 q, then (a, b θ p) < (a, b θ q) If p> 1 q and θ is nonnull under a then (a, b θ p) Â (a, b θ q) To grasp the meaning of the next axiom, let Y T Θ Then given a A and b 00, b 0 B such that b 00 Â b 0, a betony conditional on T isabetb satisfying b (θ) = b 00 (θ) if θ Y and b (θ) = b 0 (θ) if θ T The conditional replacement axiom below requires that for every given action, if the decision maker is indifferent between betting on Y conditional on T and betting on the outcome of a coin flip with probability of winning α conditional on T, then he must remain indifferent when the constant utility bets that figure in the bet on Y conditional on T change, and/or when the lotteries on any other effect change Formally, (A10) (Conditional replacement) For any a A and Y T Θ, if a, b (θ) b (θ) b (θ) θ Y α b (θ)+(1 α) b (θ) θ Y θ T Y a, α b (θ)+(1 α) b (θ) θ T Y θ Θ T b (θ) θ Θ T 15
16 then, for all b 0, b 00 B such that b 00 Â b 0 and b B, a, b0 (θ) b (θ) b00 (θ) θ Y α b 00 (θ)+(1 α) b 0 (θ) θ T Y a, α b 00 (θ)+(1 α) b 0 (θ) θ Θ T b (θ) θ Y θ T Y θ Θ T 43 Action-dependent subjective probabilities without expected utility The following theorem asserts that a preference relation is a continuous weak order satisfying conditional monotonicity, and conditional replacement and is additively separable on A B if and only if there exist a unique family of action-dependent subjective probabilities, {µ (a) a A}, on Θ and utility functions, V on B and v on A such that, for any action-bet pair ³ (a, b) C, (a, b) V Σ θ Θ µ (θ; a) b θ b (θ) + v (a), where b θ b (θ) denotes the element of B whose θ coordinate is b (θ) The following matrix is useful in trying to understand the meaning of the expression Σ θ Θ µ (θ; a) b θ b (θ) Given a bet b, and θ i Θ, b θi b (θ i ) B is 16
17 depicted in the i th column of the following matrix: Θ θ 1 θ 2 θ n θ 1 b1 (θ 1 )=b (θ 1 ) b2 (θ 1 ) bn (θ n ) θ 2 b1 (θ 2 ) b2 (θ 2 )=b (θ 2 ) bn (θ 2 ) θ n b1 (θ n ) b2 (θ n ) b n (θ n )=b (θ n ) A real-valued function, f, on a convex subset, S, of a linear space is mixture continuous if f (αp +(1 α) q) is continuous in α for all p, q S Theorem 3 Let < be a nondegenerate preference relation on C and suppose that the richness assumption holds, then (a) The following conditions are equivalent: (ai) < satisfies (A1), (A2), (A9) and (A10) and its restriction to A B satisfies (A3) (A5) (aii) There exist a family of probability measures {µ (a) a A} on Θ, continuous strictly monotonic, real valued function V on B and a continuous real-valued v on A such that, V and v are jointly cardinal and, for all (a, b), (a 0,b 0 ) C, (a, b) < (a 0,b 0 ) 17
18 if and only if V ³ Σ θ Θ µ (θ; a) b θ b (θ) + v (a) V ³ Σ θ Θ µ (θ; a 0 ) b θ b 0 (θ) + v (a 0 ) (3) (b) Thefamilyofprobabilitymeasures{µ (a) a A} on Θ is unique and µ (θ; a) =0if and only if θ is null given a 44 Probabilistically sophisticated choice Like all theories based on the analytical framework of Savage (1954), the models of probabilistically sophisticated choice of Machina and Schmeidler (1992, 1995), and Grant (1995) assume, implicitly, that constant acts are constant-utility acts Adding this assumption to the model of the preceding section yields a model of probability sophisticated choice with action-dependent subjective probabilities Note that in this case Z (θ) =Z for all θ Θ Let B c := {b B b (θ) =b (θ 0 ) θ, θ 0 Θ} be the set of constant bets Theorem 4 Let < be a nondegenerate preference relation on C and suppose that the richness assumption holds, then (a) The following conditions are equivalent: (ai) < satisfies (A1), (A2), (A9) and (A10) and its restriction to A B c satisfies (A3) (A5) 18
19 (aii) There exist a family of probability measures {µ ( ; a) a A} on Θ, acontinuous, strictly monotonic, real-valued function, V, on (Z) and a continuous real valued function v on A such that, V and v are jointly cardinal and for all (a, b), (a 0,b 0 ) C, Ã! Ã! X X (a, b) < (a 0,b 0 ) V µ (θ; a) b (θ) + v (a) V µ (θ; a 0 ) b 0 (θ) + v (a 0 ) (4) θ Θ θ Θ (b) Thefamilyofprobabilitymeasures{µ ( ; a) a A} on Θ is unique and µ (θ; a) =0if and only if θ is null given a 5 Concluding Remarks Using the analytical framework of Karni (2005) and invoking the device of roulette lotteries, this paper examines the choice-based foundations of subjective probabilities and decisionmaking under uncertainty The choice set consists of actions-bet pairs, where bets are lottery-valued functions on a set of effects and actions represent initiatives that decision makers believe might affect the likely realization of different effects The decision makers preferences over actions and bets reveal these beliefs as well as their risk preferences The main results of this paper are: (a) A general subjective expected utility theory with action-dependent subjective probabilities and effect-dependent risk preferences The cases of effect-independent risk preferences is obtained as special instance (b) A nonexpected utility theory involving well-defined families of action-dependent sub- 19
20 jective probabilities on effects The utility assigned to action-bet pairs depends on the convex combination of constant-utility bets corresponding to the coordinates of the original bet, where the weights are the action-dependent subjective probabilities A probabilistic sophistication version of this model is obtained as a special case in which constant bets are the constant-utility bets (c) As argued in Karni (2005a), the canonical probabilities defined in these models constitute a numerical representation of a decision maker s degree of belief regarding the likely realization of events conditional on his choice of action 6 Proofs 61 Proof of Theorem 1 (ai) (aii) The restriction of < to A B is a continuous weak order satisfying axioms (A3) (A5) Hence there exist continuous, jointly cardinal, real-valued functions V on B and v on A such that the restriction of < to A B has an additive representation, a, b 7 V b + v (a) (see Wakker 1989, Theorem III41) Let b, b B satisfy a, b  a, b (that such bets exist is an implication of nondegeneracy of < and the richness assumption) Invoking the uniqueness properties of V and v normalize these functions so that V b =0 and V b =1 20
21 Axiom (A2) implies the Archimedean axiom Hence, by the von Neumann Morgenstern theorem, < satisfies (A1), (A2) and (A6) if there exist jointly cardinal, continuous functions {w a (,θ):z (θ) R θ Θ} such that, for every given a A, and all b, b 0 B, (a, b) < (a, b 0 ) X θ Θ w a (b (θ),θ) X θ Θ w a (b 0 (θ),θ), (5) where w a (b (θ),θ)= P z Z(θ) w a (z, θ) b (z,θ) If θ is null given a then w a (,θ) is a constant function Invoking the joint cardinality normalize the additive-valued functions {w a (,θ)} θ Θ as follows: Let w a b (θ),θ = v (a) / Θ, for all θ Θ and a A, and P θ Θ wa b (θ),θ (v (a) / Θ ) = 1 (Note that if θ is null given a then w a (,θ)=v (a) / Θ ) Because < is nondegenerate Axiom (A7) implies that, for all a, a 0 A and θ Θ (a; <) Θ (a 0 ; <), there are numbers β (θ; a, a 0 ) > 0 and γ (θ; a, a 0 ) such that w a (,θ)=β (θ; a, a 0 ) w a 0 (,θ)+γ (θ; a, a 0 ) (6) Probabilities and utilities: Define π (θ; a) =w a b (θ);θ v (a) Θ for all θ Θ and a A (7) Then, by the normalization of w a, for all a A, π (θ; a) 0 for all θ Θ,and P θ Θ π (θ; a) = 1 But w a ( ; θ) =β (θ; a, a 0 ) w a 0 ( ; θ)+γ (a, a 0,θ) and, by the normalization of w a,forevery 21
22 a A, w a b (θ),θ = v (a) / Θ, for all θ Θ Hence Equations (7) and (8) imply that β (θ; a, a 0 )= γ (a, a 0,θ)= v (a) Θ β (θ; a, a0 ) v (a0 ) Θ (8) π (θ; a) π (θ; a 0 ) for all a, a0 A and θ Θ satisfying π (θ; a 0 ) > 0 (9) For any given p (Z (θ)), θ Θ and a A, define U (p; θ, a) = w a (p; θ) v (a) Θ /π (θ; a) if π (θ; a) > 0 and U (p; θ, a) =ū otherwise (10) Note that, for all a A and θ Θ (a 0 ; <) Θ (a; <), h i h i w a U (p; θ, a 0 0 (p; θ) v(a0 ) w Θ a (p; θ) v(a) Θ )= = π (θ; a 0 ) β (a,a,θ)π (θ; a 0 ) 0 = h i w a (p; θ) v(a) Θ π (θ; a) = U (p; θ, a), (11) where the third inequality is implied by equation (9) Hence U (p; θ, a) :=U (p; θ) for all a A and θ Θ (a; <) Note that U b (θ),θ =1and U b (θ),θ =0for all θ Θ By definition, w a (p; θ) =π (θ; a) U (p; θ)+ v (a) Θ for all a A, θ Θ and p (Z (θ)) (12) Hence X w a (b (θ),θ)=x U (b (θ),θ) π (θ; a)+v (a) (13) θ Θ θ Θ By von Neumann-Morgenstern theorem, U (,θ) if affine Hence, by the standard argument, there is a real-valued function, u, on C such that U (b (θ),θ)= P z Z(θ) u (z,θ) b (z, θ) 6 6 See Kreps (1988) 22
23 Next observe that, by the uniqueness properties of the additive representations and the normalizations, for every a A and b B, X X u (z,θ) b (z, θ) π (θ; a)+v (a) =V b + v (a) (14) θ Θ z Z(θ) For every given b B, let b ³ δ z( b), where δ z( b) = δ z(θ; b(θ)) (that is, z θ; b (θ) is the θ Θ certainty equivalent of b (θ) in θ) Then δ z ( b) B and, because B is convex, z θ; b (θ) = λ (θ) z θ 0 ; b (θ 0 ),λ(θ) > 0, θ Θ Hence X X u (z,θ) b (z, θ) π (θ; a) = X u λ (θ) z θ 0 ; b (θ 0 ),θ π (θ; a) θ Θ (15) θ Θ z Z(θ) But equations (14) and (15) imply that P θ Θ u z θ; b (θ),θ π (θ; a) is independent of a Hence u λ (θ) z θ 0 ; b (θ 0 ),θ = u λ (θ 00 ) z θ 0 ; b (θ 0 ),θ 00 for all θ, θ 00 Θ (a; <) Thus, by equation (14), u λ (θ) z θ 0 ; b (θ 0 ),θ = V b for all θ Θ (a; <) Representation of <: For all (a, b) and (a 0,b 0 ) in C, (a, b) < (a 0,b 0 ) if and only if X X u (z, θ) b (z,θ) X u (z,θ) b 0 (z,θ) π (θ; a 0 )+v(a 0 ) π (θ; a)+v (a) X θ Θ θ Θ z Z(θ) z Z(θ) To see this observe that, by the richness assumption, there are constant utility bets b and b 0 such that a, b (a, b) < (a 0,b 0 ) a 0, b 0 By transitivity, a, b < a 0, b 0 But a, b < a 0, b 0 if and only if V b + v (a) V b0 + v (a 0 ) And, by equation (14) and (15), (16) V b + v (a) V b0 + v (a 0 ) if and only if X X u (z, θ) b (z,θ) π (θ; a)+v (a) X X θ Θ θ Θ z Z(θ) 23 z Z(θ) u (z,θ) b 0 (z,θ) π (θ; a 0 )+v(a 0 ) (17)
24 But a, b (a, b) if and only if P θ Θ w P a b (θ),θ = θ Θ w a (b (θ),θ) Thus, by equations and (13), X a, b (a, b) X and θ Θ (a 0,b 0 ) a 0, b 0 X θ Θ z Z(θ) X z Z(θ) u (z,θ) b (z, θ) π (θ; a) = X X θ Θ z Z(θ) u (z, θ) b 0 (z, θ) π (θ; a 0 )= X X θ Θ z Z(θ) u (z,θ) b (z, θ) π (θ; a), (18) u (z,θ) b 0 (z,θ) π (θ; a 0 ) Equation (16) follows from equations (17) (19) and transitivity This completes the proof that (ai) (aii) (19) (aii) (ai) That (aii) implies (A1) (A2) is immediate That (aii) implies (A6) follows from the von Neumann Morgenstern theorem The proof of (A7) is immediate That (A3) (A5) are implied by Wakker (1989) Theorem III41 The functions (V,v) are jointly cardinal and u z θ; b (θ),θ = V b for all θ Θ Moreover, by the richness assumption, for every z Z (θ) there is b θ δ z B such that u (z,θ) =V b θ δ z (To see this let λ (θ 0 ) z satisfy u (λ (θ 0 ) z,θ 0 )=u (z,θ) and let b θ δ z = δλ(θ)z θ Θ ) Thus u and v are jointly cardinal (b) To prove the uniqueness of {π ( ; a)} a A suppose, by way of negation, that for some a A there exists a probability measure, µ ( ; a), on Θ and µ ( ; a) 6= π ( ; a) Then there are effects θ 0,θ 00 Θ such that µ (θ 0 ; a) >π(θ 0 ; a) and π (θ 00 ; a) >µ(θ 00 ; a) Without essential loss of generality suppose that µ (θ; a) = π (θ; a) for all θ Θ {θ 0,θ 00 } (Note 24
25 that µ (θ; a) >π(θ; a) and π (θ 0 ; a) >µ(θ 0 ; a) imply that θ and θ 0 are nonnull given a) Let b B and r 0,r 00 R such that u (r 0,θ 0 ) > P z Z(θ) u(z,θ0 ) b (z; θ 0 ) and u (r 00,θ 00 ) < P ³ z Z(θ 00 ) u(z, θ00 ) b (z; θ 00 ) satisfy b θ 0δ r 0 δ θ 00 r 00 b Then the representation implies that u (r 0,θ 0 ) π (θ 0 ; a)+u (r 00,θ 00 ) π (θ 00 ; a) =V b [π (θ 0 ; a)+π (θ 00 ; a)] and, since π (θ 0 ; a)+π (θ 00 ; a) =µ (θ 0 ; a)+µ (θ 00 ; a), u (r 0,θ 0 ) µ (θ 0 ; a)+u (r 00,θ 00 ) µ (θ 00 ; a) >V b [µ (θ 0 ; a)+µ (θ 00 ; a)] A contradiction Hence µ ( ; a) =π ( ; a) By definition π(θ; a) =0for all θ that is null given a 62 Proof of Theorem 2 (ai) (aii) By Theorem 1, for all a A, θ, θ 0 Θ (a; <), and p, q (Z (θ)) (Z (θ 0 )), (a, b θ p) < (a, b θ q) X u (z; θ) p (z) X u (z; θ) q (z) (20) z Z(θ) z Z(θ) and (a, b θ 0p) < (a, b θ 0q) X u (z; θ 0 ) p (z) X u (z; θ 0 ) q (z) (21) z Z(θ 0 ) z Z(θ 0 ) By axiom (A8) X u (z; θ) p (z) X u (z; θ) q (z) X u (z; θ 0 ) p (z) X u (z; θ 0 ) q (z) (22) z Z(θ) z Z(θ) z Z(θ 0 ) z Z(θ 0 ) 25
26 Let u ( ; θ 0 ):=u ( ), then, by the uniqueness of the von Neumann Morgenstern utility function, for all a A and θ Θ (a; <), u ( ; θ) =σ (θ) u ( )+κ (θ),σ(θ) > 0 (23) But, by Theorem 1, u (λ (θ) z,θ) =u (λ (θ 0 ) z,θ 0 ) for all a A and θ, θ 0 Θ (a; <) Hence, by equation (23),for all z, [σ (θ) σ (θ 0 )] u (λ (θ) z,θ) =κ (θ 0 ) κ (θ) (24) But u ( ; θ) is non-constant Hence σ (θ) σ (θ 0 )=κ(θ 0 ) κ (θ) =0 Invoking the uniqueness of the von Neumann Morgenstern utility function, let σ (θ) =1and κ (θ) =0for all θ Θ Thus, by equation (23), u ( ; θ) =u ( ) That u and v are continuous and jointly cardinal is implied by Theorem 1 This complete the proof that (ai) implies (aii) The proof that (aii) implies axioms (A1) (A7) follows from Theorem 1 The proof that it also implies axiom (A8) is straightforward The proof of part (b) follows from that of part (b) in Theorem 1 63 Proof of Theorem 3 (ai) (aii) The restriction of < to A B is a continuous weak order satisfying axioms (A3) (A5) Hence there exist continuous, jointly cardinal, real-valued functions V on B and v on A such that the restriction of < to A B has an additive representation, a, b 7 V b + v (a) (see Wakker 1989, Theorem III41) 26
27 By the richness assumption, for all (a, b), (a 0,b 0 ) C, thereexist b, b 0 B such that a, b (a, b) and a 0, b 0 (a 0,b 0 ) Transitivity implies that, for all (a, b), (a 0,b 0 ) C, (a, b) < (a 0,b 0 ) a, b < a 0, b 0 V b + v (a) V b0 + v (a 0 ) (25) Because < is nondegenerate axioms (A2) and (A9) imply that for every given (a, b) C there exists a unique number ϕ (a, b) [0, 1], defined by ϕ (a, b) =Sup{ϕ (a, b) < a, ϕ b +(1 ϕ) b }, (26) such that (a, b) a, ϕ (a, b) b +(1 ϕ (a, b)) b (For a proof see Kreps (1988) using the fact that (A2) implies the Archimedean axiom) Moreover, by the convexity of B, ϕ (a, b) b +(1 ϕ (a, b)) b B Clearly, by axioms (A2) and (A9), ϕ(a, ) is mixture continuous and monotonic with respect to first order stochastic dominance For every a A and θ Θ define µ (θ; a) by a, b θ b (θ) (a, µ (θ; a) b +(1 µ (θ; a)) b ) (27) By the preceding discussion µ (θ; a) [0, 1], is well-defined, and µ (θ; a) =0if and only if θ is nullthis proves part(b) of the theorem Fix a A and, without loss of generality, let θ 1 Θ (a; <) (that is, θ 1 is nonnull under a) Consider two constant utility bets, b 00 and b 0, such that b 00 Â b 0 For every i =1,,n 1 and a A let α i (a) be given by a, b00 (θ 1 ),, b 00 (θ i ), b 0 (θ i+1 ),, b 0 (θ n ) a, α i (a) b 00 +(1 α i (a)) b 0 (28) 27
28 Then α i (a) [0, 1] is well defined and, by axiom (A10) with T = Θ, is independent of the constant utility bets b 00 and b 0 assigned to the subset of effects {θ 1,, θ i } and {θ i+1,, θ n } provided that b 00 Â b 0 Let α 0 (a) =0and define τ i (a) =α n 1 (a) α n 2 (a) α i (a)(1 α i 1 (a)), i =1,,n 1, τ n (a) =(1 α n 1 (a)) ThenΣ n i=1τ i (a) =1 Note that b θ1 b (θ 1 )(θ 1 )=b (θ 1 ) Hence by repeated application of axiom (A10), for any (a, b) C, a, b (θ 1 ) on θ 1 b (θ 2 ) on θ 2 b (θ 3 ) on θ 3 b (θ 4 ) on θ 4 b (θ n ) on θ n a, α 1 (a) b (θ 1 )+(1 α 1 (a))b θ2 b (θ 2 )(θ 1 ) on θ 1 α 1 (a) b θ1 b (θ 1 )(θ 2 )+(1 α 1 (a))b (θ 2 ) on θ 2 b (θ 3 ) on θ 3 b (θ 4 ) on θ 4 b (θ n ) on θ n 28
29 a, α 2 (a) h α 1 (a) b θ1 b (θ 1 )+(1 α 1 (a))b θ2 b (θ 2 ) i (θ 1 )+(1 α 2 (a)) b θ3 b (θ 3 )(θ 1 ) on θ 1 α 2 (a) h α 1 (a) b θ1 b (θ 1 )+(1 α 1 (a))b θ1 b (θ 2 ) i (θ 2 )+(1 α 2 (a)) b θ3 b (θ 3 )(θ 2 ) on θ 2 α 2 (a) h α 1 (a) b θ1 b (θ 1 )+(1 α 1 (a)) b θ2 b (θ 2 ) i (θ 3 )+(1 α 2 (a)) b (θ 3 ) on θ 3 b (θ 4 ) on θ 4 b(θ n ) on θ n a, τ 1 (a) b (θ 1 )+τ 2 (a) b θ2 b (θ 2 )(θ 1 )+ + τ n (a) b θn b (θ n )(θ 1 ) on θ 1 τ 1 (a) b θ1 b (θ 1 )(θ 2 )+τ 2 (a) b (θ 2 )+ + τ n (a) b θn b (θ n )(θ 2 ) on θ 2 τ 1 (a) b θ1 b (θ 1 )(θ n )+τ 2 (a) b θ2 b (θ 2 )(θ n )+ + τ n (a) b (θ n ) on θ n = ³ a, ³ τ 1 (a) b θ1 b (θ 1 )+τ 2 (a) b θ2 b (θ 2 )+ + τ n (a) b θn b (θ n ) = ³ a, Σ n i=1τ i (a) b θi b (θ i ) Thus any action-bet pair (a, b) is indifferent to the action-bet pair a, b where b B is a convex combination of the constant utility bets corresponding to the coordinates of the original bet b But ³ a, b θ j b (a, τ j (a) b +(1 τ j (a)) b ), (29) 29
30 ³ Hence τ j (a) =µ (θ j ; a) for all θ j,j=1,, n, and a A, and (a, b) a, Σ θ Θ µ (θ; a) b θ b (θ) By (25), for all (a, b), (a 0,b 0 ) C, (a, b) < (a 0,b 0 ) V ³ Σ θ Θ µ (θ; a) b θ b (θ) + v (a) V ³ Σ θ Θ µ (θ; a 0 ) b 0 θ b0 (θ) + v (a 0 ) (30) This completes the proof that (i) (aii) The proof that (aii) (ai) is immediate 64 Proof of Theorem 4 (ai) (aii) Since constant bets are constant utility bets, b θ p =[p,, p] for all θ Θ Identify [p,, p] with p and, with slight abuse of notation, let V ([p,, p]) = V (p) Then P θ Θ µ (θ; a) b θp =[p,, p] and, by Theorem 3 and the identification above, for all (a, b), (a 0,b 0 ) C, (a, b) < (a 0,b 0 ) V Ã! X µ (θ; a) b (θ) + v (a) V θ Θ Ã! X µ (θ; a) b 0 (θ) + v (a 0 ) (31) θ Θ The rest of the proof of the theorem follows from the proof of Theorem 3 30
31 References [1] Anscombe, F and R Aumann, (1963) A definition of Subjective Probability, Annals of Mathematical Statistics 34, [2] Grant, S (1995) Subjective Probability without Monotonicity: Or How Machina s Mom May also Be Probabilistically Sophisticated," Econometrica 63, [3] Karni, E (1993) A Definition of Subjective Probabilities with State-Dependent Preferences Econometrica 61, [4] Karni, E (2005) Subjective Expected Utility Theory without States of the World, Journal of Mathematical Economics forthcoming [5] Karni, E (2005a) On the Existence and Uniqueness of Subjective Probabilities Representing Beliefs, Unpublished paper [6] Kreps, D (1988) Notes on the Theory of Choice Westview Press, Boulder [7] Machina, M J and D Schmeidler (1992) A More Robust Definition of Subjective Probability, Econometrica 60, [8] Machina, M J and D Schmeidler, (1995) "Bayes without Bernoulli: Simple Conditions for Probabilistically Sophisticated Choice" Journal of Economic Theory, 67, [9] Ramsey, Frank P (1931) Truth and Probability, In The Foundations of Mathematics and Other Logical Essays London: K Paul, Trench, Truber and Co 31
32 [10] Savage, L J (1954) The Foundations of Statistics New York: John Wiley and Sons [11] Wakker, P P (1989) The Additive Representations of Preferences Dordrecht: Kluwer Academic Publishers 32
Action-Independent Subjective Expected Utility without States of the World
heoretical Economics Letters, 013, 3, 17-1 http://dxdoiorg/10436/tel01335a004 Published Online September 013 (http://wwwscirporg/journal/tel) Action-Independent Subjective Expected Utility without States
More informationA Theory of Bayesian Decision Making
A Theory of Bayesian Decision Making Edi Karni November 11, 2009 Abstract This paper presents a complete, choice-based, axiomatic Bayesian decision theory. It introduces a new choice set consisting of
More informationA theory of Bayesian decision making with action-dependent subjective probabilities
Econ Theory 2011 48:125 146 DOI 10.1007/s00199-010-0542-1 RESEARCH ARTICLE A theory of Bayesian decision making with action-dependent subjective probabilities Edi Karni Received: 11 November 2009 / Accepted:
More informationReverse Bayesianism: A Generalization
Reverse Bayesianism: A Generalization Edi Karni Johns Hopkins University and Warwick Business School Quitzé Valenzuela-Stookey Northwestern University Marie-Louise Vierø Queen s University December 10,
More informationA Reformulation of the Maxmin Expected Utility. Model with Application to Agency Theory
A Reformulation of the Maxmin Expected Utility Model with Application to Agency Theory Edi Karni Johns Hopkins University November 16, 2006 Abstract Invoking the parametrized distribution formulation of
More informationAxiomatic Foundations of Expected Utility. and Subjective Probability
Axiomatic Foundations of Expected Utility and Subjective Probability Edi Karni Johns Hopkins University April 29, 2013 This is a chapter commissioned for the Handbook of Risk and Uncertainty (Machina and
More informationRecursive Ambiguity and Machina s Examples
Recursive Ambiguity and Machina s Examples David Dillenberger Uzi Segal May 0, 0 Abstract Machina (009, 0) lists a number of situations where standard models of ambiguity aversion are unable to capture
More informationA Theory of Bayesian Decision Making
A Theory of Bayesian Decision Making Edi Karni September 7, 2009 Abstract This paper presents a complete, choice-based, axiomatic Bayesian decision theory. It introduces a new choice set consisting of
More informationMathematical Social Sciences
Mathematical Social Sciences 74 (2015) 68 72 Contents lists available at ScienceDirect Mathematical Social Sciences journal homepage: www.elsevier.com/locate/econbase Continuity, completeness, betweenness
More informationDominance and Admissibility without Priors
Dominance and Admissibility without Priors Jörg Stoye Cornell University September 14, 2011 Abstract This note axiomatizes the incomplete preference ordering that reflects statewise dominance with respect
More informationSubjective Probabilities on a State Space
Subjective Probabilities on a State Space Edi Karni May 24, 2011 Abstract This paper extends the analytical framework of Karni (2010) to include a state space and advances a choice-based definition of
More informationSeptember 2007, France
LIKELIHOOD CONSISTENCY M h dabd ll i (& P t P W kk ) Mohammed Abdellaoui (& Peter P. Wakker) September 2007, France A new method is presented for measuring beliefs/likelihoods under uncertainty. It will
More informationAn Axiomatic Model of Reference Dependence under Uncertainty. Yosuke Hashidate
An Axiomatic Model of Reference Dependence under Uncertainty Yosuke Hashidate Abstract This paper presents a behavioral characteization of a reference-dependent choice under uncertainty in the Anscombe-Aumann
More informationChoice under Uncertainty
In the Name of God Sharif University of Technology Graduate School of Management and Economics Microeconomics 2 44706 (1394-95 2 nd term) Group 2 Dr. S. Farshad Fatemi Chapter 6: Choice under Uncertainty
More informationChapter 2. Decision Making under Risk. 2.1 Consequences and Lotteries
Chapter 2 Decision Making under Risk In the previous lecture I considered abstract choice problems. In this section, I will focus on a special class of choice problems and impose more structure on the
More informationJournal of Mathematical Economics
Journal of Mathematical Economics 63 (2016) 164 173 Contents lists available at ScienceDirect Journal of Mathematical Economics journal homepage: www.elsevier.com/locate/jmateco A theory of stochastic
More informationTwo-Stage-Partitional Representation and Dynamic Consistency 1
Two-Stage-Partitional Representation and Dynamic Consistency 1 Suguru Ito June 5 th, 2015 Abstract In this paper, we study an individual who faces a three-period decision problem when she accepts the partitional
More informationAn additively separable representation in the Savage framework
An additively separable representation in the Savage framework Brian Hill HEC Paris October 29, 2007 Abstract This paper elicits an additively separable representation of preferences in the Savage framework
More informationAmbiguity Aversion: An Axiomatic Approach Using Second Order Probabilities
Ambiguity Aversion: An Axiomatic Approach Using Second Order Probabilities William S. Neilson Department of Economics University of Tennessee Knoxville, TN 37996-0550 wneilson@utk.edu April 1993 Abstract
More informationFamiliarity Breeds Completeness
Familiarity Breeds Completeness Edi Karni February 20, 2013 Abstract This is a study of the representations of subjective expected utility preferences that admits state-dependent incompleteness, and subjective
More informationRecursive Ambiguity and Machina s Examples
Recursive Ambiguity and Machina s Examples David Dillenberger Uzi Segal January 9, 204 Abstract Machina (2009, 202) lists a number of situations where Choquet expected utility, as well as other known models
More informationAmbiguity under Growing Awareness
Ambiguity under Growing Awareness Adam Dominiak 1 and Gerelt Tserenjigmid 2 1,2 Department of Economics, Virginia Tech September 10, 2018 Abstract In this paper, we study choice under growing awareness
More informationSecond-Order Expected Utility
Second-Order Expected Utility Simon Grant Ben Polak Tomasz Strzalecki Preliminary version: November 2009 Abstract We present two axiomatizations of the Second-Order Expected Utility model in the context
More informationGreat Expectations. Part I: On the Customizability of Generalized Expected Utility*
Great Expectations. Part I: On the Customizability of Generalized Expected Utility* Francis C. Chu and Joseph Y. Halpern Department of Computer Science Cornell University Ithaca, NY 14853, U.S.A. Email:
More informationPre-probabilistic sophistication. Osamu Kada. July, 2009
Pre-probabilistic sophistication Osamu Kada Faculty of Engineering, Yokohama National University 79-5 Tokiwadai, Hodogaya, Yokohama 240-850, Japan (E-mail: smkada@ynu.ac.jp) July, 2009 We consider pre-probabilistic
More informationNotes on Supermodularity and Increasing Differences. in Expected Utility
Notes on Supermodularity and Increasing Differences in Expected Utility Alejandro Francetich Department of Decision Sciences and IGIER Bocconi University, Italy March 7, 204 Abstract Many choice-theoretic
More informationAre Probabilities Used in Markets? 1
Journal of Economic Theory 91, 8690 (2000) doi:10.1006jeth.1999.2590, available online at http:www.idealibrary.com on NOTES, COMMENTS, AND LETTERS TO THE EDITOR Are Probabilities Used in Markets? 1 Larry
More informationRECURSIVE AMBIGUITY AND MACHINA S EXAMPLES 1. INTRODUCTION
INTERNATIONAL ECONOMIC REVIEW Vol. 56, No., February 05 RECURSIVE AMBIGUITY AND MACHINA S EXAMPLES BY DAVID DILLENBERGER AND UZI SEGAL University of Pennsylvania, U.S.A.; Boston College, U.S.A., and Warwick
More informationAnscombe & Aumann Expected Utility Betting and Insurance
Anscombe & Aumann Expected Utility Betting and Insurance Econ 2100 Fall 2017 Lecture 11, October 3 Outline 1 Subjective Expected Utility 2 Qualitative Probabilities 3 Allais and Ellsebrg Paradoxes 4 Utility
More informationLecture Notes 1: Decisions and Data. In these notes, I describe some basic ideas in decision theory. theory is constructed from
Topics in Data Analysis Steven N. Durlauf University of Wisconsin Lecture Notes : Decisions and Data In these notes, I describe some basic ideas in decision theory. theory is constructed from The Data:
More informationExpected Utility Framework
Expected Utility Framework Preferences We want to examine the behavior of an individual, called a player, who must choose from among a set of outcomes. Let X be the (finite) set of outcomes with common
More informationThis corresponds to a within-subject experiment: see same subject make choices from different menus.
Testing Revealed Preference Theory, I: Methodology The revealed preference theory developed last time applied to a single agent. This corresponds to a within-subject experiment: see same subject make choices
More information3 Intertemporal Risk Aversion
3 Intertemporal Risk Aversion 3. Axiomatic Characterization This section characterizes the invariant quantity found in proposition 2 axiomatically. The axiomatic characterization below is for a decision
More informationReverse Bayesianism : A Choice-Based Theory of Growing Awareness
Reverse Bayesianism : A Choice-Based Theory of Growing Awareness Edi Karni Johns Hopkins University Marie-Louise Vierø Queen s University December 11, 212 Abstract This paper introduces a new approach
More informationExpected utility without full transitivity
Expected utility without full transitivity Walter Bossert Department of Economics and CIREQ University of Montreal P.O. Box 6128, Station Downtown Montreal QC H3C 3J7 Canada FAX: (+1 514) 343 7221 e-mail:
More informationA MODEL OF PROCEDURAL DECISION MAKING IN THE PRESENCE OF RISK 1. INTRODUCTION
INTERNATIONAL ECONOMIC REVIEW Vol. 43, No. 4, November 2002 A MODEL OF PROCEDURAL DECISION MAKING IN THE PRESENCE OF RISK BY JUAN DUBRA AND EFE A. OK 1 Department of Economics, New York University, Universidad
More informationProbability Elicitation for Agents with Arbitrary Risk Preferences
Probability Elicitation for Agents with Arbitrary Risk Preferences Nicolas S. Lambert July 2018 Abstract A principal asks an agent for his probability assessment of a random event, and wants to induce
More information14.12 Game Theory Lecture Notes Theory of Choice
14.12 Game Theory Lecture Notes Theory of Choice Muhamet Yildiz (Lecture 2) 1 The basic theory of choice We consider a set X of alternatives. Alternatives are mutually exclusive in the sense that one cannot
More informationSubjective Probabilities on a State Space
Subjective Probabilities on a State Space Edi Karni Johns Hopkins University July 27, 2009 Abstract This paper extends the work of Karni (2009) in two distinct directions. First, it generalizes the model
More informationPreference, Choice and Utility
Preference, Choice and Utility Eric Pacuit January 2, 205 Relations Suppose that X is a non-empty set. The set X X is the cross-product of X with itself. That is, it is the set of all pairs of elements
More informationIncomplete Preferences under Uncertainty: Indecisiveness in Beliefs vs. Tastes
Incomplete Preferences under Uncertainty: Indecisiveness in Beliefs vs. Tastes Efe A. Ok Pietro Ortoleva Gil Riella April 7, 2011 Abstract We investigate the classical Anscombe-Aumann model of decision-making
More informationRecitation 7: Uncertainty. Xincheng Qiu
Econ 701A Fall 2018 University of Pennsylvania Recitation 7: Uncertainty Xincheng Qiu (qiux@sas.upenn.edu 1 Expected Utility Remark 1. Primitives: in the basic consumer theory, a preference relation is
More informationCoherent Choice Functions Under Uncertainty* OUTLINE
Coherent Choice Functions Under Uncertainty* Teddy Seidenfeld joint work with Jay Kadane and Mark Schervish Carnegie Mellon University OUTLINE 1. Preliminaries a. Coherent choice functions b. The framework
More informationIntertemporal Risk Aversion, Stationarity, and Discounting
Traeger, CES ifo 10 p. 1 Intertemporal Risk Aversion, Stationarity, and Discounting Christian Traeger Department of Agricultural & Resource Economics, UC Berkeley Introduce a more general preference representation
More informationBayesian consistent prior selection
Bayesian consistent prior selection Christopher P. Chambers and Takashi Hayashi August 2005 Abstract A subjective expected utility agent is given information about the state of the world in the form of
More informationTransitive Regret over Statistically Independent Lotteries
Transitive Regret over Statistically Independent Lotteries April 2, 2012 Abstract Preferences may arise from regret, i.e., from comparisons with alternatives forgone by the decision maker. We show that
More informationA Theory of Subjective Compound Lotteries
A Theory of Subjective Compound Lotteries Haluk Ergin Washington University in St Louis and Faruk Gul Princeton University September 2008 Abstract We develop a Savage-type model of choice under uncertainty
More informationHomework #6 (10/18/2017)
Homework #6 (0/8/207). Let G be the set of compound gambles over a finite set of deterministic payoffs {a, a 2,...a n } R +. A decision maker s preference relation over compound gambles can be represented
More informationFamiliarity breeds completeness
Familiarity breeds completeness Edi Karni Economic Theory ISSN 0938-2259 Volume 56 Number 1 Econ Theory (2014) 56:109-124 DOI 10.1007/s00199-013-0777-8 1 23 Your article is protected by copyright and all
More informationBayesian Updating for General Maxmin Expected Utility Preferences
Bayesian Updating for General Maxmin xpected Utility Preferences Marciano Siniscalchi September 14, 2001 First draft Comments welcome! Abstract A characterization of generalized Bayesian updating in a
More informationExpected Utility Preferences for Contingent Claims and Lotteries
Expected Utility Preferences for Contingent Claims and Lotteries Felix Kubler University of Zurich Swiss Finance Institute Larry Selden Columbia University University of Pennsylvania Xiao Wei University
More informationINTENSITY OF THE SENSE OF FAIRNESS: MEASUREMENT AND BEHAVIORAL CHARACTERIZATION. October 2000
INTENSITY OF THE SENSE OF FAIRNESS: MEASUREMENT AND BEHAVIORAL CHARACTERIZATION Edi Karni Department of Economics Johns Hopkins University Zvi Safra Faculty of Management Tel Aviv University October 2000
More informationWorst-Case Expected Utility
Worst-Case Expected Utility Shiri Alon Abstract The paper presents a model in which a decision maker, having a preference relation over purely subjective acts, deviates minimally from the Subjective Expected
More informationIntertemporal Risk Aversion, Stationarity and Discounting
Intertemporal Risk Aversion, Stationarity and Discounting Job Market Paper, November 2007 Christian P. TRAEGER Department of Agricultural & Resource Economics, UC Berkeley Department of Economics, UC Berkeley
More informationWORKING PAPER SERIES
INTERNATIONAL CENTRE FOR ECONOMIC RESEARCH WORKING PAPER SERIES Paolo Ghirardato and Massimo Marinacci RISK, AMBIGUITY, AND THE SEPARATION OF UTILITY AND BELIEFS Working Paper no. 21/2001 September 2001
More informationCompleting the State Space with Subjective States 1
Journal of Economic Theory 105, 531539 (2002) doi:10.1006jeth.2001.2824 Completing the State Space with Subjective States 1 Emre Ozdenoren Department of Economics, University of Michigan, Ann Arbor, Michigan
More informationEvent-Separability in the Ellsberg urn
Econ Theory (2011) 48:425 436 DOI 10.1007/s00199-011-0652-4 SYMPOSIUM Event-Separability in the Ellsberg urn Mark J. Machina Received: 5 April 2011 / Accepted: 9 June 2011 / Published online: 16 July 2011
More informationCautious and Globally Ambiguity Averse
Ozgur Evren New Economic School August, 2016 Introduction Subject: Segal s (1987) theory of recursive preferences. Ambiguity aversion: Tendency to prefer risk to ambiguity. (Ellsberg paradox) risk objective,
More informationMeasurable Ambiguity. with Wolfgang Pesendorfer. August 2009
Measurable Ambiguity with Wolfgang Pesendorfer August 2009 A Few Definitions A Lottery is a (cumulative) probability distribution over monetary prizes. It is a probabilistic description of the DMs uncertain
More informationChoice under uncertainty
Choice under uncertainty Expected utility theory The agent chooses among a set of risky alternatives (lotteries) Description of risky alternatives (lotteries) a lottery L = a random variable on a set of
More informationProjective Expected Utility
Projective Expected Utility Pierfrancesco La Mura Department of Microeconomics and Information Systems Leipzig Graduate School of Management (HHL) April 24, 2006 1 Introduction John von Neumann (1903-1957)
More informationLexicographic Expected Utility with a Subjective State Space
Lexicographic Expected Utility with a Subjective State Space Youichiro Higashi Kazuya Hyogo August 25, 2008 Abstract This paper provides a model that allows for a criterion of admissibility based on a
More informationCautious Expected Utility and the Certainty Effect
Cautious Expected Utility and the Certainty Effect Simone Cerreia-Vioglio David Dillenberger Pietro Ortoleva May 2013 Abstract One of the most prominently observed behavioral patterns in decision making
More informationA Theory of Subjective Learning
A Theory of Subjective Learning David Dillenberger Juan Sebastián Lleras Philipp Sadowski Norio Takeoka July 2014 Abstract We study an individual who faces a dynamic decision problem in which the process
More informationGame Theory without Decision-Theoretic Paradoxes
Game Theory without Decision-Theoretic Paradoxes Pierfrancesco La Mura Department of Economics and Information Systems HHL - Leipzig Graduate School of Management Jahnallee 59, 04109 Leipzig (Germany)
More informationDECISIONS UNDER UNCERTAINTY
August 18, 2003 Aanund Hylland: # DECISIONS UNDER UNCERTAINTY Standard theory and alternatives 1. Introduction Individual decision making under uncertainty can be characterized as follows: The decision
More informationProbabilistic Subjective Expected Utility. Pavlo R. Blavatskyy
Probabilistic Subjective Expected Utility Pavlo R. Blavatskyy Institute of Public Finance University of Innsbruck Universitaetsstrasse 15 A-6020 Innsbruck Austria Phone: +43 (0) 512 507 71 56 Fax: +43
More informationWhat Are Asset Demand Tests of Expected Utility Really Testing?
What Are Asset Demand Tests of Expected Utility Really Testing? Felix Kubler, Larry Selden and Xiao Wei August 25, 2016 Abstract Assuming the classic contingent claim setting, a number of nancial asset
More informationPreferences for Randomization and Anticipation
Preferences for Randomization and Anticipation Yosuke Hashidate Abstract In decision theory, it is not generally postulated that decision makers randomize their choices. In contrast, in real life, even
More informationAmbiguity and Nonexpected Utility
Ambiguity and Nonexpected Utility Edi Karni, Fabio Maccheroni, Massimo Marinacci October 24, 2013 We gratefully thank Pierpaolo Battigalli, Veronica R. Cappelli, Simone Cerreia-Vioglio, and Sujoy Mukerji
More informationChoice Under Uncertainty
Choice Under Uncertainty Z a finite set of outcomes. P the set of probabilities on Z. p P is (p 1,...,p n ) with each p i 0 and n i=1 p i = 1 Binary relation on P. Objective probability case. Decision
More informationLecture notes on statistical decision theory Econ 2110, fall 2013
Lecture notes on statistical decision theory Econ 2110, fall 2013 Maximilian Kasy March 10, 2014 These lecture notes are roughly based on Robert, C. (2007). The Bayesian choice: from decision-theoretic
More informationCorrelated Equilibria of Classical Strategic Games with Quantum Signals
Correlated Equilibria of Classical Strategic Games with Quantum Signals Pierfrancesco La Mura Leipzig Graduate School of Management plamura@hhl.de comments welcome September 4, 2003 Abstract Correlated
More informationRelative Benefit Equilibrating Bargaining Solution and the Ordinal Interpretation of Gauthier s Arbitration Scheme
Relative Benefit Equilibrating Bargaining Solution and the Ordinal Interpretation of Gauthier s Arbitration Scheme Mantas Radzvilas July 2017 Abstract In 1986 David Gauthier proposed an arbitration scheme
More informationExtensions of Expected Utility Theory and some Limitations of Pairwise Comparisons
Extensions of Expected Utility Theory and some Limitations of Pairwise Comparisons M. J. SCHERVISH Carnegie Mellon University, USA T. SEIDENFELD Carnegie Mellon University, USA J. B. KADANE Carnegie Mellon
More informationRisk Aversion over Incomes and Risk Aversion over Commodities. By Juan E. Martinez-Legaz and John K.-H. Quah 1
Risk Aversion over Incomes and Risk Aversion over Commodities By Juan E. Martinez-Legaz and John K.-H. Quah 1 Abstract: This note determines the precise connection between an agent s attitude towards income
More informationEndogenizing Prospect Theory s Reference Point. by Ulrich Schmidt and Horst Zank
Endogenizing Prospect Theory s Reference Point by Ulrich Schmidt and Horst Zank No. 1611 March 2010 Kiel Institute for the World Economy, Hindenburgufer 66, 24105 Kiel, Germany Kiel Working Paper No. 1611
More informationGame Theory. Lecture Notes By Y. Narahari. Department of Computer Science and Automation Indian Institute of Science Bangalore, India August 2012
Game Theory Lecture Notes By Y. Narahari Department of Computer Science and Automation Indian Institute of Science Bangalore, India August 2012 Chapter 7: Von Neumann - Morgenstern Utilities Note: This
More informationA CONCEPTUAL FOUNDATION FOR THE THEORY OF RISK AVERSION YONATAN AUMANN. Bar Ilan University Ramat Gan, Israel
A CONCEPTUAL FOUNDATION FOR THE THEORY OF RISK AVERSION YONATAN AUMANN Bar Ilan University Ramat Gan, Israel Abstract. Classically, risk aversion is equated with concavity of the utility function. In this
More informationNorthwestern University
Northwestern University 00 Sheridan Road 580 Leverone Hall Evanston, IL 6008-04 USA Discussion Paper #54 April 0, 0 Preference for Randomization and Ambiguity Aversion Key words: Ambiguity; randomization,
More informationSubjective multi-prior probability: a representation of a partial l
Subjective multi-prior probability: a representation of a partial likelihood relation TAU, 22-11-2011 Outline of the talk: Motivation Background Examples Axioms Result Related literature Outline of the
More informationThree contrasts between two senses of coherence
Three contrasts between two senses of coherence Teddy Seidenfeld Joint work with M.J.Schervish and J.B.Kadane Statistics, CMU Call an agent s choices coherent when they respect simple dominance relative
More informationECO 317 Economics of Uncertainty Fall Term 2009 Notes for lectures 3. Risk Aversion
Reminders ECO 317 Economics of Uncertainty Fall Term 009 Notes for lectures 3. Risk Aversion On the space of lotteries L that offer a finite number of consequences (C 1, C,... C n ) with probabilities
More informationA Bayesian Approach to Uncertainty Aversion
Review of Economic Studies (2005) 72, 449 466 0034-6527/05/00190449$02.00 c 2005 The Review of Economic Studies Limited A Bayesian Approach to Uncertainty Aversion YORAM HALEVY University of British Columbia
More informationCOMPARATIVE STATICS FOR RANK-DEPENDENT EXPECT- ED UTILITY THEORY
COMPARATIVE STATICS FOR RANK-DEPENDENT EXPECT- ED UTILITY THEORY John Quiggin Department of Agricultural and Resource Economics University of Maryland Quiggin, J. (1991), Comparative statics for Rank-Dependent
More informationAxiomatic Decision Theory
Decision Theory Decision theory is about making choices It has a normative aspect what rational people should do... and a descriptive aspect what people do do Not surprisingly, it s been studied by economists,
More informationPh.D. Preliminary Examination MICROECONOMIC THEORY Applied Economics Graduate Program June 2016
Ph.D. Preliminary Examination MICROECONOMIC THEORY Applied Economics Graduate Program June 2016 The time limit for this exam is four hours. The exam has four sections. Each section includes two questions.
More informationSubjective recursive expected utility
Economic Theory 2005) DOI 10.1007/s00199-005-0041-y RESEARCH ARTICLE Peter Klibanoff Emre Ozdenoren Subjective recursive expected utility Received: 23 November 2004 / Accepted: 7 September 2005 Springer-Verlag
More informationCALIFORNIA INSTITUTE OF TECHNOLOGY
DIVISION OF THE HUMANITIES AND SOCIAL SCIENCES CALIFORNIA INSTITUTE OF TECHNOLOGY PASADENA, CALIFORNIA 91125 RANGE CONVEXITY AND AMBIGUITY AVERSE PREFERENCES. Paolo Ghirardato California Institute of Technology
More informationA Rothschild-Stiglitz approach to Bayesian persuasion
A Rothschild-Stiglitz approach to Bayesian persuasion Matthew Gentzkow and Emir Kamenica Stanford University and University of Chicago December 2015 Abstract Rothschild and Stiglitz (1970) represent random
More informationDecision-making with belief functions
Decision-making with belief functions Thierry Denœux Université de Technologie de Compiègne, France HEUDIASYC (UMR CNRS 7253) https://www.hds.utc.fr/ tdenoeux Fourth School on Belief Functions and their
More informationCautious Expected Utility and the Certainty Effect
Cautious Expected Utility and the Certainty Effect Simone Cerreia-Vioglio David Dillenberger Pietro Ortoleva February 2014 Abstract Many violations of the Independence axiom of Expected Utility can be
More informationComment on The Veil of Public Ignorance
Comment on The Veil of Public Ignorance Geoffroy de Clippel February 2010 Nehring (2004) proposes an interesting methodology to extend the utilitarian criterion defined under complete information to an
More informationEcon 2148, spring 2019 Statistical decision theory
Econ 2148, spring 2019 Statistical decision theory Maximilian Kasy Department of Economics, Harvard University 1 / 53 Takeaways for this part of class 1. A general framework to think about what makes a
More informationRationality and Uncertainty
Rationality and Uncertainty Based on papers by Itzhak Gilboa, Massimo Marinacci, Andy Postlewaite, and David Schmeidler Warwick Aug 23, 2013 Risk and Uncertainty Dual use of probability: empirical frequencies
More informationNash Equilibrium without. Mutual Knowledge of Rationality 1. Kin Chung Lo. Department of Economics, University oftoronto, July, 1995.
Nash Equilibrium without Mutual Knowledge of Rationality 1 Kin Chung Lo Department of Economics, University oftoronto, Toronto, Ontario, Canada M5S 1A1 July, 1995 Abstract This paper denes an equilibrium
More informationA Rothschild-Stiglitz approach to Bayesian persuasion
A Rothschild-Stiglitz approach to Bayesian persuasion Matthew Gentzkow and Emir Kamenica Stanford University and University of Chicago January 2016 Consider a situation where one person, call him Sender,
More informationExpressive Rationality
Expressive Rationality Larry Blume Cornell University & The Santa Fe Institute & IHS, Vienna What Caused the Financial Crisis? Blume Expressive Rationality 1 What Caused the Financial Crisis? MATH! Blume
More informationQuantum Decision Theory
Quantum Decision Theory V.I. Yukalov and D. Sornette Department of Management, Technology and Economics\ ETH Zürich Plan 1. Classical Decision Theory 1.1. Notations and definitions 1.2. Typical paradoxes
More informationNo-envy in Queueing Problems
No-envy in Queueing Problems Youngsub Chun School of Economics Seoul National University Seoul 151-742, Korea and Department of Economics University of Rochester Rochester, NY 14627, USA E-mail: ychun@plaza.snu.ac.kr
More information