Luce Rule with Attraction System

Size: px
Start display at page:

Download "Luce Rule with Attraction System"

Transcription

1 Luce Rule with Attraction System Jiangtao Li Rui Tang April 26, 206 Abstract We develop a two-stage extension of Luce s random choice model that incorporates an attraction parameter. Each option is associated with an attraction value and a Luce value. Attraction value measures the option s ability to attract the agent and Luce value measures the option s desirability. In the first stage, the agent is randomly attracted by some alternative according to an attraction-valued Luce-type formula. The agent then chooses among alternatives that are comparable with this option according to another Luce-valued Luce-type formula. While retaining the simplicity of the Luce model, the theory can explain the attraction (violations of regularity), compromise, similarity effects, and violations of stochastic transitivity, which are well-documented behavioral phenomena in individual choice. We are grateful to Soo Hong Chew, Faruk Gul, Shaowei Ke, Matthew Kovach, John Quah and Satoru Takahashi for helpful discussions. Part of this paper was written while Jiangtao Li was visiting the Economics Department at University of Michigan and he would like to thank the institution for its hospitality and support. All remaining errors are our own. Preliminary and Incomplete. Comments are Welcome. Department of Economics, National University of Singapore, jasonli07@gmail.com Department of Economics, National University of Singapore, a0292@u.nus.edu

2 Introduction Deterministic behavior is often assumed in models of economic choice. However, in both experimental and market settings, individual choice responses typically exhibit variability; see, for example, Sippel (997), McFadden (200) and Manzini, Mariotti, and Mittone (200). In this paper, we study a random choice model and assume choice responses to be given by a probability distribution ρ that indicates the probability ρ(a, A) that when the possible options faced by the agent are the alternatives in A, some alternative in A is chosen, as in the seminal work of Luce (959), Tversky (972b) and more recently, Gul, Natenzon, and Pesendorfer (204) (henceforth, GNP), Manzini and Mariotti (204), Echenique, Saito, and Tserenjigmid (204) and Brady and Rehbeck (205). In the celebrated Luce model, each option x is associated with a Luce value v x that measures the option s desirability. The probability of choosing x from an option set A containing x is ρ({x}, A) : v x / v y. y A Despite many attractive features, there are well-documented violations. We develop a twostage extension of Luce s random choice model to address such violations. The novel aspect of the paper is that we incorporate an attraction parameter into the Luce model. Each option x is associated with an attraction value χ x and a Luce value v x. Attraction value measures the option s ability to attract the agent. As in the Luce model, Luce value measures the option s desirability. In the first stage, the agent is randomly attracted by some alternative according to an attraction-valued Luce-type formula. The agent then chooses among alternatives that are comparable with this option according to another Luce-valued Luce-type formula. This type of decision procedure is ubiquitous in daily life. An investor is attracted by a particular car manufacturer (possibly due to prior knowledge). But when she compares within the car manufacturing industry, she might eventually invest in another car manufacturer, which has stronger fundamentals. A shopper is attracted by a recently launched handbag of a luxury brand (possibly through advertisement). But when she visits the store and compares the various handbags, she might eventually purchase a different handbag, which fits her even better. A foodie is attracted by a spicy dish at a Michelin restaurant (possibly recommended by a friend). But when she dines there and compares the various choices from the menu, she might eventually order a non-spicy dish, which suits her flavor even better. 2

3 While retaining the simplicity of the Luce rule, the theory can explain the attraction (violations of regularity), compromise, similarity effects, and violations of stochastic transitivity, which are well-documented behavioral phenomena in individual choice. what follows, we shall illustrate how we apply the theory to explain the attraction effect (violations of regularity). A fuller discussion of the attraction effect and the other behavioral phenomena can be found in Section 4. Most random choice models including Luce rule (Luce (959)), elimination-by-aspects rule (Tversky (972b)), attribute rule (GNP (204)) and random consideration set rule (Manzini and Mariotti (204)) obey the regularity property (or monotonicity). That is, when an alternative is added to the option set, the choice probability of the original members of the choice set cannot increase. Although regularity seems innocuous, this property conflicts with empirical evidence from the marketing literature, most notably Huber, Payne, and Puto (982) and Simonson (989). The following is a slight modification of an example proposed by Huber, Payne, and Puto (982). Example. Let S {x, y, z}. Assume that x, y and z are three brands of beer. Subjects know only the price and the average quality ratings. See Figure. Brand Dimension : Price Dimension 2: Quality x Low Medium y High High z Medium Low We might have ρ({x}, {x, y}) and ρ({x}, {x, y, z}) 3. Thus, we have a violation of 3 8 regularity. In Example, since z is worse than x in both dimensions, it is likely that x and z are comparable. Also assume that y and z are not comparable, since y is better in one dimension but worse in the other dimension. On the one hand, the addition of z decreases the probability that the agent is attracted by x and y in the first stage. On the other hand, when the agent is attracted by z, she still considers x, but not y in the second stage. Furthermore, it is plausible that x has a higher Luce value than z, since it is more desirable in both dimensions. If z has a sufficiently high attraction value and x has a sufficiently high Luce value relative to z, then the addition of z increases the probability of x being chosen. Hence, our model allows for violations of the regularity property. 3 In

4 Preference x Dimension : Price z y 0 Dimension 2: Quality Preference Figure : Placement of asymmetrically dominated decoys. Luce model can be viewed as a special case of Luce rule with attraction system, where all options are comparable. No matter which option the agent is attracted by in the first stage, she still considers all options in the second stage and eventually chooses an option using the Luce-valued Luce-type formula. This paper joins a growing literature that studies random choice. The closest to our model in terms of representation is the attribute rule (GNP (204)). GNP reinterpret the options as bundles of attributes. In the attribute rule, the agent is randomly attracted by some attribute according to a Luce-type formula and only options that has that attribute are considered in the second stage. Consider an investor who is choosing from car manufacturer x and several other companies. If we add another car manufacturer y in the option set, this never increases the probability of x being considered in the second stage. At best, y does not introduce attributes that are non-existent in the original option set, and the probability of x being considered in the second stage remains unchanged. In our model, the agent is randomly attracted by some option rather than some attribute. For example, suppose there are two car manufacturers x, y and other companies, an investor need not be fixated to investing in any particular industry. Instead, she is attracted by each company with certain probability, possibly due to prior knowledge of the companies, transaction history of the company stocks, or recommendations by a financial advisor. When she is attracted by either x or y, she compares all car manufacturers when she eventually makes the investment decision. Therefore, the addition of y increases the probability of x being considered in the second stage. 4

5 The rest of the paper is organized as follows. Section 2 introduces the axiom of additivity of relative probabilities and shows that Luce rule is the only choice rule that satisfies additivity of relative probabilities. Section 3 presents Luce rule with attraction system. Luce rule with attraction system weakens additivity of relative probabilities. Theorem 2, our main result, shows that Luce rule with attraction system is the only choice rule that satisfies the weak version of additivity of relative probabilities and an additional transitivity axiom. Section 4 applies the theory to explain the attraction (violations of regularity), compromise, similarity effects, and violations of stochastic transitivity. 2 Preliminaries Let X be a nonempty finite set of choice options and denote by A 2 X its power set. Let A + A\. We use A, B, C,... to denote option sets and x, y, z,... to denote options. With slight abuse of notation, sometimes we write x for {x}. To simplify the statements below, we use the following notational convention: AB : A B, xa : {x} A, xy : {x} {y}. Definition. A random choice rule is a map ρ : A A + [0, ] such that for all A A +, i) ρ(, A) is additive; ii) ρ(a, A) ; and iii) ρ({x}, A) (0, ] for all x A. The interpretation is that ρ(a, A) denotes the probability that when the possible options faced by the agent are the alternatives in A, some alternative in A is chosen. Additivity is the requirement that ρ(, A) is a probability. Equation ρ(a, A) is the feasibility constraint; ρ must choose among options available in A. We write ρ(x A) rather than ρ({x}, A) and ρ(a A) rather than ρ(a, A). Luce rule. The Luce rule (Luce (959)) is a well-known behavioral optimization model that can be described as follows. Let v : A R ++ and v x : v({x}). Such a function v is a 5

6 Luce value if it is additive. Hence, v( ) 0, and for all A A +, v(a) x A v x. Call the choice rule ρ a Luce rule if there exists a Luce value v such that ρ(x A) v x /v(a) whenever x A A. Clearly, every Luce value v represents a unique choice rule. Definition 2. In an option set C that contains A and B, we define the relative probability of choosing options in A and B conditional on C as follows: For notational simplicity, we write β(a, B C) ρ(a C) ρ(b C). β(a, B) β(a, B AB) ρ(a AB) ρ(b AB). We introduce the following axiom of additivity of relative probabilities, which asserts that for option x and option set A disjoint, the relative probability of choosing options in xa and B conditional on xab is the sum of two components: the relative probability of choosing x and B conditional on xb and the relative probability of choosing options in A and B conditional on AB. Axiom A - Additivity of Relative Probabilities : If x / A, then β(xa, B) β(x, B)+ β(a, B). It is easy to see that additivity of relative probabilities is another probabilistic form of independence from irrelevant alternatives. The following theorem shows that Luce rule is the only choice rule that satisfies additivity of relative probabilities. Theorem. A choice rule satisfies additivity of relative probabilities if and only if it is a Luce rule. 6

7 3 Luce rule with attraction system Each option is associated with an attraction value and a Luce value. An attraction value is a function χ : A R ++ and measures the option s ability to attract the agent. Again, we write χ x rather than χ({x}). Like Luce value, we require χ to be additive. Hence, χ( ) 0, and for all A A +, χ(a) : χ x. x A For a partition of X and x X, denote by X x the element of the partition that contains x. Call the choice rule ρ a Luce rule with attraction system if there exists a partition (X i ) n i of X, an attraction value χ and a Luce value v such that ρ(x A) χ(a X x) χ(a) v(x) v(a X x ) (a) whenever x A A. We call ((X i ) n i, χ, v) an attraction system. We say that the attraction system ((X i ) n i, χ, v) represents ρ if equation (a) holds for all such x, A. Clearly, every attraction system represents a unique choice rule. We say that two options x and y are comparable if the relative probability of choosing x and y is menu independent; that is, the relative probability of choosing x and y is the same in any option set that contains x and y. As in GNP (204), whether two options are comparable is subjective and must be derived from behavior. It cannot be decided based on physical characteristics of the options and therefore, it is a property of the choice rule and not of the options. More formally, Definition 3. Two options x and y are comparable if in any option set A that contains x and y, we have β(x, y A) β(x, y). We write x y if x and y are comparable and x y if otherwise. The relation is symmetric and reflexive. Also write A B if x A, y B, we have x y. Say that an option set A is a collection of comparable options if x, y A x y. In this case, we write A L. Note that even if A L and B L, it need not be the case that AB L. As the next example demonstrates, the relation is not necessarily transitive. Example 2. Let X {x, y, z} and the random choice rule ρ is given by ρ(x xy) ρ(y xy) 2, 7

8 ρ(y yz) ρ(z yz) 2, ρ(x xyz) ρ(y xyz) ρ(z xyz) 3, ρ(x xz) 3, ρ(z xz) 2 3. It is straightforward to verify that x y, y z and yet x z. Our first axiom asserts that the relation is transitive; that is, if x and y are comparable, y and z are comparable, then x and z are comparable. Axiom T - Transitivity : The relation is transitive. Next we define the notion of overlap of two options sets: when A and B have elements in common, they overlap. Even if A and B have no elements in common, they overlap if there are elements of A and B that are comparable. Definition 4. A and B are non-overlapping if x A, y B, x y. We write A B if A and B are non-overlapping. We introduce the following weak version of additivity of relative probabilities. It requires additivity of relative probability only for the case where B is a collection of comparable options, and xa and B are non-overlapping. More formally, Axiom WA - Weak Additivity of Relative Probabilities : If x / A and xa B L, then β(xa, B) β(x, B) + β(a, B). We show that Luce rule with attraction system is the only choice rule that satisfies transitivity and weak additivity of relative probabilities. Theorem 2. A choice rule satisfies transitivity and weak additivity of relative probabilities if and only if it is a Luce rule with attraction system. 4 Discussion Despite many attractive features of the Luce model, there are well-documented violations; see Rieskamp, Busemeyer, and Mellers (2006) for an excellent survey. In this section, we We provide further discussions on this axiom in Appendix A.4. 8

9 apply our theory to explain the attraction (violations of regularity), compromise, similarity effects, and violations of stochastic transitivity. When all options are comparable, Luce rule with attraction system reduces to the Luce model. Consequently, the data cannot exhibit any violations of the Luce model. For simplicity of exposition, in what follows, we shall exclude the discussion of such case Attraction effect (violations of regularity) As discussed in the introduction, the attraction effect can be described as follows: ρ(x xyz) > ρ(x xy). In this subsection, we show that Luce rule with attraction system allows for attraction effect (violation of regularity). Let us first revisit Example. Example 3 (Example Continued). In this case, x and z are comparable; that is, x z, x y. The attraction values are χ x χ z, χ y 2. The Luce values are v x 3, v y v z. We have exhibiting attraction effect. ρ(x xyz) and ρ(x xy) χ x + χ z v x 3 χ x + χ y + χ z v x + v z 8, χ x χ x + χ y 3, Proposition below provides a necessary and sufficient condition on the attraction system for attraction effect. Proposition. The following two statements are equivalent: i) ρ(x xyz) > ρ(x xy); ii) x z, x y and v x χ y χ z > χ x v z (χ x + χ y + χ z ). Proof. ii) i) is straightforward. For i) ii), it suffices to consider the following cases. Case i) x y, x z. We have a contradiction, since ρ(x xyz) v x χ x + χ y χ x + χ y + χ z v x + v y 2 Note that it is never revealed from the choice data that for any two options x and y, x y. Suppose to the contrary, x y for any x and y. Then for any option set A that contains options z and w, we have β(z, w A) β(z, w) χz χ w ; that is, z w. We have a contradiction. 9

10 Case ii) x z, x y. In this case, < v x v x + v y ρ(x xy). ρ(x xyz) > ρ(x xy) χ x + χ z v x > χ x + χ y + χ z v x + v z χ x χ x + χ y v x χ y χ z > χ x v z (χ x + χ y + χ z ). Case iii) x y, y z. We have a contradiction, since ρ(x xyz) χ x χ x + χ y + χ z < χ x χ x + χ y ρ(x xy). 4.2 Compromise effect The compromise effect is another observed violation of the Luce model; see, for example, Simonson (989) and Simonson and Tversky (992). Consider three options x, y and z. x and z are extreme options where x is the highest in both quality and price, z is the lowest in both quality and price, whereas y has moderate values on both dimensions. The agent s option set is {x, y} in Experiment and {x, y, z} in Experiment 2. The experimental data shows that the probability of choosing y increases when moving from Experiment to 2. This pattern is called the compromise effect because y may be viewed as a compromise between the two extreme options x and z. As in Rieskamp, Busemeyer, and Mellers (2006), compromise effect could be described as follows: β(x, y xyz) < β(x, y). In this subsection, we show that Luce rule with attraction system allows for compromise effect. Consider the following example. 0

11 Example 4. Let v x v y v z χ z and χ x χ y 2 where x z, x y. We have exhibiting compromise effect. β(x, y xyz) (χ x + χ z ) vx v x+v z χ y 3 4, and β(x, y) χ x χ y, Proposition 2 below provides a necessary and sufficient condition on the attraction system for compromise effect. Proposition 2. The following two statements are equivalent: i) β(x, y xyz) < β(x, y); ii) x y, x z, χ x χ y and v x (χ x + χ z ) < χ y (v x + v z ); or x y, y z, χ x χ y and χ x (v y + v z ) < v y (χ y + χ z ). Proof. ii) i) is straightforward. For i) ii), it suffices to consider the following cases. and and Case i) x y, x z. We have a contradiction, since Case ii) x z, x y. In this case, x y β(x, y xyz) β(x, y). β(x, y xyz) < Case iii) x y, y z. In this case, β(x, y) χ x χ y, (χ x + χ z ) vx v x+v z χ y < v x (χ x + χ z ) < χ y (v x + v z ). β(x, y) χ x χ y, β(x, y xyz) < χ x < (χ y + χ z ) vy v y+v z χ x (v y + v z ) < v y (χ y + χ z ).

12 4.3 Similarity effect In Luce model, the addition of a third option z will take from options x and y in proportion to their original shares. More formally, β(x, y xyz) β(x, y). Similarity effect says that an additional option takes disproportionately more share from those similar to it than from dissimilar options. Suppose that three alternatives are such that x and z are somehow very similar to each other, and clearly distinct from y. This setup is discussed by Tversky (972a), building on an example of Debreu (960). Similarity effect can be formalized as follows: β(x, y xyz) < β(x, y). In this subsection, we show that the Luce rule with attraction system allows for similarity effect. Consider the following example. Example 5. Let v x v y v z χ y χ z and χ x 2 where x z, x y. We have exhibiting similarity effect. β(x, y xyz) (χ x + χ z ) vx v x+v z χ y 3 2, and β(x, y) χ x χ y 2, There are two possible explanations for the similarity effect. We provide the intuition for one explanation here. Assume that x and z are comparable, and x and y are non-overlapping. The addition of z increases the probability of x being considered in the second stage. But if z has a sufficiently high Luce value, it may be that x is eventually chosen with a lower probabily. The exact effect on the relative probability of x and y when we add a third option z depends on the comparison of two values: the attraction value of z relative to x, and the Luce value of z relative to x. Proposition 3 below provides a necessary and sufficient condition on the attraction system for similarity effect. Proposition 3. The following two statements are equivalent: i) β(x, y xyz) < β(x, y); ii) x z, x y and χz χ x or x y, y z and χz χ y < vz v x ; > vz v y. Proof. ii) i) is straightforward. For i) ii), it suffices to consider the following cases. 2

13 Case i) x y, y z. We have a contradiction, since Case ii) x z, x y. In this case, x y β(x, y xyz) β(x, y). β(x, y xyz) < β(x, y) Case iii) x y, y z. In this case, (χ x + χ z ) vx v x+v z χ y χ z χ x < v z v x. < χ x χ y β(x, y xyz) < β(x, y) χ x (χ y + χ z ) vy v y+v z χ z χ y > v z v y. < χ x χ y If similar options must have the same attraction values and Luce values, it follows from Proposition 3 that choices do not exhibit similarity effect in our model. 4.4 Stochastic intransitivity Several psychologists, starting from Tversky (969), have argued that choices may fail to be transitive. When choice is stochastic, there are many ways to define analogues of transitive behavior in deterministic models. A weak such analogue is the following: Definition 5 (Weak Stochastic Transitivity). For any x, y, z A, β(x, y) and β(y, z) imply β(x, z). While the Luce rule must necessarily satisfy weak stochastic transitivity, we show that the Luce rule with attraction system allows for violations of weak stochastic transitivity. Consider the following example. Example 6. Let v x v y v z χ x and χ y χ z 2 where x y, y z. We have β(x, y) v x v y, 3

14 but also violating weak stochastic transitivity. β(y, z) χ y χ z, β(x, z) χ x χ z 2, Proposition 4 below provides a necessary and sufficient condition on the attraction system for the violations of weak stochastic transitivity. Proposition 4. The following statements are equivalent: i) β(x, y), β(y, z), β(x, z) < ; ii) x y, y z, v x v y and χ y χ z > χ x ; or x z, x y, v x < v z and χ x χ y χ z ; or x y, y z, v y v z and χ z > χ x χ y. Proof. ii) i) is straightforward. For i) ii), it suffices to consider the following cases. Case i) x y, y z. In this case, Case ii) x z, x y. In this case, Case iii) x y, y z. In this case, β(x, y) v x v y, β(y, z) χ y χ z. β(x, z) < χ x < χ z. β(x, y) χ x χ y, β(y, z) χ y χ z. β(x, z) < v x < v z. β(x, y) χ x χ y, β(y, z) v y v z. β(x, z) < χ x < χ z. 4

15 Under reasonable conditions, the Luce rule with attraction system satisfies weak stochastic transitivity. For example, if the ordering of the attraction values χ weakly agrees with the ordering of the Luce values v. 3 In this case, it follows from Proposition 4 that choices are weakly stochastically transitive. A Appendix A. Preparatory Lemmas Lemma. If A B, then for any C, we have β(a, B ABC) β(a, B). Proof. Since A B, for any x A, y B, we have x y. For any C and x A, we have β(b, x ABC) β(y, x ABC) y B y B β(y, x AB) β(b, x AB), () where the first equality follows from the definition of β and the second equality follows from the definition of the relation. Furthermore, we have β(a, B ABC) β(x, B ABC) x A x A β(x, B AB) where the second equality follows from (). β(a, B), Lemma 2. If A B L, then β(a, B) x A β(x, B). Proof. Without loss of generality, write A {x, x 2,..., x k }. Let A A, A 2 A \{x },..., A k A k \{x k } {x k }. Repeatedly applying Axiom WA, we have β(a, B) β(x, B) + β(a 2, B) 3 More formally, v x v y v z and χ x χ y χ z. 5

16 β(x, B) + β(x 2, B) + β(a 3, B) x A β(x, B). Lemma 3. If A B and AB C L, then β(ab, C) β(a, C) + β(b, C). Proof. By Lemma 2, β(ab, C) β(x, C) x AB β(x, C) + β(x, C) x A x B β(a, C) + β(b, C). Lemma 4. If a, b, c (0, ) and a + b a b a a c c + b b c c, then Proof. Rearranging gives a b c c. a 2 c 2 + 2abc 2 + b 2 c 2 2a 2 c 2abc + a 2 (c(a + b) a)) 2 0. Therefore, a b c c. 6

17 Lemma 5. For A, B, C L and pairwise disjoint, if A B, A C and B C, then β(a, B ABC) β(a, B). Proof. By Lemma 3, we have Rearranging gives β(ab, C) β(a, C) + β(b, C); β(ac, B) β(a, B) + β(c, B); β(bc, A) β(b, A) + β(c, A). β(ab, C) β(a, C) + β(b, C) β(c, A) + β(c, B) β(bc, A) β(b, A) + β(ac, B) β(a, B). (2) By Lemma 4, we have 4 β(a, B ABC) β(a, B). Lemma 6. If a, b, c, d > 0, and then Proof. Rearranging gives Simplifying further, we have and ad bc. ( a + b + c + d )( a + c ad bc. + + ), b d (a + b + c + d)(abc + abd + abd + bcd) (a + b)(c + d)(a + c)(b + d). a 2 d 2 2abcd + b 2 c 2 (ad bc) Let ρ(a ABC) a, ρ(b ABC) b, ρ(a AB) c. 7

18 A.2 Proof of Theorem It is easy to see that Luce rule satisfies additivity of relative probabilities. In what follows, we show that if a choice rule ρ satisfies additivity of relative probabilities, then it is a Luce rule. For x / A, β(xa, B) β(x, B) + β(a, B). Using similar techniques as in the proof of Lemma 2, Lemma 3 and Lemma 5, we can show that for any A, B, C A + where A B, β(ab, C) β(a, C) + β(b, C) and β(a, B ABC) β(a, B). Therefore, it is a Luce rule. A.3 Proof of Theorem 2 We first show that if a choice rule ρ satisfies transitivity and weak additivity of relative probabilities, then ρ is a two-stage Luce rule with attraction system. With transitivity, we can partition the set X in the following way: {X i } n i, such that () X i L, for any i, 2,..., n ; and (2) X i X j for any i, j, 2,..., n and i j. Without loss of generality, we assume that there are at least two elements in the partition. 5 Consider any option set A A + and option x A. Let A i A X i, we have A n i(a X i ) A A 2 A n. It is easy to see that () A i L, for any i, 2,..., n ; and (2) A i A j for any i, j, 2,..., n and i j. Without loss of generality, we assume that x A. By definition of the relation, {x} A. Therefore, ρ(x A) ρ(x A A 2 A n ) ρ(a A A 2 A n )β(x, A A A 2 A n ) ρ(a A A 2 A n )β(x, A A ) ρ(a A A 2 A n )ρ(x A ), (3) where the third equality follows from Lemma. Now we show that β(a i, A A A 2 A n ) β(a i, A ) in a series of lemmas. Say an n n matrix M is symmetrically reciprocal (SR) if a for all i, 2,..., n and a ij a ji for all i, j, 2,..., n and i j. Say an n n matrix M is transitive if all entry of M is positive and a ij a ik a kj for all i, j, k, 2,..., n. 5 For the case where X L, it is easy to see that our model reduces to the standard Luce rule. 8

19 Lemma 7. For any SR and transitive matrix M (a ij ) n n, there exists w i, i, 2,..., n such that a ij w i w j and n w i. Proof. We explicitly construct w i for i, 2,..., n. Let w i i a i n i a, i, 2,..., n. i It is easy to see that For i > and j >, a i w i w and a j w wj. a i,j a i a j w i w w wj w i wj. Let N (,,..., ) T denote the n vector of s. Lemma 8. For any SR and transitive matrices M (a ij ) n n and M (ā ij ) n n, if N (M M) 0, we have a ij b ij for all i, j, 2,..., n. Proof. By Lemma 7, we can choose w i for M and w i for M such that a ij w i w j, ā ij w i w j, n w i, i n i w i. Since N (M M 2 ) 0, we have (N (M M)) j (N M) j (N M)j n (a kj ā kj ) k n k ( w k w j w k w j ) Therefore, w j w j and a ij ā ij. w j w j 0. 9

20 Lemma 9. For distinct partition elements A i, A j, A k, we have β(a i, A j A i A j A k ) β(a i, A j ). Proof. This follows from Lemma 5. Lemma 0. For distinct partition elements A i, A j, A k, we have β(a i, A j )β(a j, A k ) β(a i, A k ). Proof. By Lemma 9, we have β(a i, A j A i A j A k ) β(a i, A j ), β(a j, A k A i A j A k ) β(a j, A k ), β(a i, A k A i A j A k ) β(a i, A k ). By the definition of β, β(a i, A k A i A j A k ) β(a i, A j A i A j A k )β(a j, A k A i A j A k ), we have β(a i, A k ) β(a i, A j )β(a j, A k ). Lemma. β(a i, A j ) β(a i, A j A). Proof. By Lemma 2, we have β(a i, A i ) β(x, A i ) β(a j, A i ). (4) x A i By the definition of β, Therefore, it follows from (4) and (5) that j:j i β(a i, A i ) j:j i β(a j, A i A). (5) j:j i β(a j, A i ) j:j i β(a j, A i A). (6) Now let a ji β(a j, A i ) and ā ji β(a j, A i A) if j i. Let a ii ā ii. Therefore, it follows from the (6) that a ji j j ā ji. (7) Furthermore, by Lemma 0, a ji a ik a jk. By the definition of β, ā ji ā ik ā jk. Therefore, by Lemma 8, we have a ij ā ij. Equivalently, β(a i, A j ) β(a i, A j A). 20

21 Therefore, ρ(x A) ρ(a A A 2 A n )ρ(x A ) ρ(a A A 2 A n ) n i ρ(a i A A 2 A n ) ρ(x A ) n i β(a i, A A A 2 A n ) ρ(x A ) n i β(a i, A ) ρ(x A ). (8) Construction of the attraction value χ. Fix an arbitrary x X and let χ( x). Define χ(y) β(y, x)χ( x) β(y, x) for any y / X. Fix an arbitrary ȳ X 2. Define χ(x) β(x, ȳ)χ(ȳ) β(x, ȳ)β(ȳ, x) for any x X and x x. Lemma 2. If x X i and ỹ X j, i j, then β( x, ỹ) χ( x) χ(ỹ). Proof. By symmetry, it suffices to consider the following cases. Case i) If x / X and ỹ / X, then β( x, ỹ) β( x, ỹ x xỹ) β( x, x x xỹ)β( x, ỹ x xỹ) β( x, x)β( x, ỹ) χ( x) χ(ỹ), where the first line and third line follow from Lemma 5. Case ii) If x x X, then ỹ / X. By construction, β( x, ỹ) χ(ỹ) χ( x) χ(ỹ). Case iii) If x X, x x and ỹ / X 2, then β( x, ỹ) β( x, ỹ xỹȳ) β( x, ȳ xỹȳ)β(ȳ, ỹ xỹȳ) β( x, ȳ)β(ȳ, ỹ) χ( x) χ(ȳ) χ(ȳ) χ(ỹ) χ( x) χ(ỹ), where the first line and third line follow from Lemma 5. 2

22 Case iv) If x X, x x and ỹ ȳ X 2, then by construction, β( x, ȳ) β( x,ȳ)χ(ȳ) χ(ȳ) χ( x). χ(ȳ) Case v) If x X, x x and ỹ X 2, ỹ ȳ. Repeatedly applying Axiom WA, we have β( x x, ỹȳ) β( x, ỹȳ) + β( x, ỹȳ) β(ỹȳ, x) + β(ỹȳ, x) β(ỹ, x) + β(ȳ, x) + β(ỹ, x) + β(ȳ, x). Similarly, Since β(ỹȳ, x x) β( x, ỹ) + β( x, ỹ) + β( x, ȳ) + β( x, ȳ). β( x x, ỹȳ)β(ỹȳ, x x) ( β(ỹ, x) + β(ȳ, x) + β(ỹ, x) + β(ȳ, x) )( β( x, ỹ) + β( x, ỹ) + β( x, ȳ) + β( x, ȳ) ), by Lemma 6, we have β(ỹ, x)β(ȳ, x) β(ȳ, x)β(ỹ, x). Therefore, β( x, ỹ) β( x, ȳ)β( x, ỹ)β(ȳ, x) χ( x) χ( x) χ(ȳ) χ(ȳ) χ(ỹ) χ( x) χ( x) χ(ỹ). Lemma 3. If A X i, B X j, i j, then Proof. By Lemma 2, β(a, B) χ(a) χ(b). β(a, B) x A β(x, B) 22

23 β(b, x) x A x A y B β(y, x) χ(y) x A y B χ(x) χ(a) χ(b). By Lemma 3, ρ(x A) n i β(a i, A ) ρ(x A ) n i2 β(a i, A ) + ρ(x A ) n i2 χ(a i ) + ρ(x A ) χ(a ) χ(a ) χ(a) ρ(x A ). (9) Construction of the Luce value v. For each partition element X i, fix an arbitrary option ˆx i X i. Define v(x i ) β(x i, ˆx i ), for any x i X i. Lemma 4. If x i, x i X i and x i x i, then β(x i, x i) v(x i) v(x i). Proof. By construction, v(x i ) β(x i, ˆx i ) and v(x i) β(x i, ˆx i ). Therefore, β(x i, x i) β(x i, x i x i x iˆx i ) β(x i, ˆx i x i x iˆx i )β(ˆx i, x i x i x iˆx i ) β(x i, ˆx i )β(ˆx i, x i) v(x i) v (x i ). By Lemma 4, ρ(x A ) β(x, A ) 23

24 β(a, x) y A β(y, x) y A v(y) v(x) v(x) v(a ). Continuing from (9), ρ(x A) χ(a ) χ(a) ρ(x A ) χ(a ) χ(a) This concludes the proof for this direction. v(x) v(a ). Next we show that if a choice rule ρ is a two-stage Luce rule with attraction system, it satisfies transitivity and weak additivity of relative probabilities. If a choice rule ρ is a two-stage Luce rule with attraction system, there exists an attraction system ((X i ) n i, χ, v) such that ρ(x A) χ(a X x) χ(a) v(x) v(a X x ) (a) whenever x A A. Consider the following coarsening of the partition (X i ) n i. For each element X i, if X i >, X i remains an element of the new partition. 6 Let B 0 denote the union of the partition elements whose cardinality is one; that is, B 0 j: Xj X j. After the coarsening, relabel the elements of new partition as B 0, B, B 2,..., B m. x, y B i, Consider the new partition (B i ) n i0. Now for any partition element B i, i, if v(x) v(y) χ(x) χ(y), we union all these elements together with B 0. After such coarsening, relabel the elements of new partition as C 0, C, C 2,..., C m. For any x C 0, we define χ (x) v (x) χ(x). Otherwise, define χ (x) χ(x) and v (x) v(x). 6 For any set S, let S denote its cardinality. 24

25 We show that the attraction system ((C i ) m i0, χ, v ) also represents the choice data. Take any arbitrary x and A containing x, we need to show that ρ(x A) χ (A C x ) χ (A) v (x) v (A C x ). (0) If x C i for some i, then it is trivial to see that equality (0) holds. If x C 0, then χ (A C x ) χ (A) χ(a C x) χ(a) χ(a X x) χ(a) χ(a X x) χ(a) ρ(x A), v (x) v (A C x ) χ(x) χ(a C x ) χ(x) χ(a X x ) v(x) v(a X x ) where the first equality follows from definition of χ and v, and the third equality follows from the property of C 0. Therefore, the attraction system ((C i ) m i0, χ, v ) also represents the choice data. We show that the attraction system ((C i ) m i0, χ, v ) satisfies transitivity and weak additivity of relative probabilities. Note that if x, y C i, then x y. Therefore, to show transitivity, it suffices to show that for some x C i, whenever x y for some y, y C i. Suppose not, there exists some y x and yet y C j for some j i. Either x or y belongs to C k for some k. Without loss of generality, we assume y C k for some k. Our construction ensures that there exists z C k such that z y and Therefore, v (y) v (z) χ (y) χ (z). () β(x, y) and β(x, y xyz) χ (x) v (x) χ (x)+χ (y) v (x) χ (y) v (y) χ (x)+χ (y) v (y) χ (x) χ (y), χ (x) v (x) χ (x)+χ (y)+χ (z) v (x) χ (y)+χ (z) v (y) χ (x)+χ (y)+χ (z) v (y)+v (z) χ (x) (χ (y) + χ (z)) v (y) v (y)+v (z) 25

26 χ (x) v (y) + v (z) χ (y) + χ (z) v (y) χ (x) x (y) + x (z). χ (y) + χ (z) x (y) χ (x) χ (y) β(x, y), where the inequality follows from (). But since y x, β(x, y) β(x, y xyz). We have a contradiction. This concludes the proof that transitivity is satisfied. For weak additivity of relative probabilities, if x / A and xa B L, β(xa, B) χ(xa) χ(xa)+χ(b) χ(b) χ(xa)+χ(b) χ(x) χ(x)+χ(b) χ(b) χ(x)+χ(b) + χ(a) χ(a)+χ(b) χ(b) χ(a)+χ(b) β(x, B) + β(a, B). A.4 Weak additivity of relative probabilities Axiom WA asserts that if x / A and xa B L, then β(xa, B) β(x, B) + β(a, B). Here we consider possible weakening of this axiom, namely Axiom WA* and Axiom WA**. We present examples where the choice data satisfies Axiom T and Axiom WA* (Axiom WA** respectively), but cannot be represented by a two-stage Luce rule with attraction system. Axiom WA* : If x y and xy B L, then β(xy, B) β(x, B) + β(y, B). Axiom WA** : If x / A and xa y L, then β(xa, y) β(x, y) + β(a, y). Example 7 discusses choice data that satisfies both Axiom T and Axiom WA*, but cannot be represented by a two-stage Luce rule with attraction system. Example 7. Let X {a, b, c, d} and the random choice rule ρ is given by ρ(a abcd) ρ(b abcd) ρ(c abcd) 3 0, ρ(d abcd) 0 ρ(x xyz) 3 ρ(x xy) 2 for distinct x, y, z {a, b, c, d} for distinct x, y {a, b, c, d} 26

27 From the choice data, a b, b c, c a and abc d. It is easy to verify that the choice data satisfies Axiom T and Axiom WA*. However, 9 β(abc, d) β(a, d) + β(bc, d) 3. The choice data does not satisfy weak additivity of relative probability and hence cannot represented by a two-stage Luce rule with attraction system. Example 8 discusses choice data that satisfies both Axiom T and Axiom WA**, but cannot be represented by a two stage Luce rule with attraction system. Example 8. Let X {a, b, c, d} and the random choice rule ρ is given by ρ(a abcd) ρ(b abcd) 6, ρ(c abcd) ρ(d abcd) 3 ρ(x xyz) 3 ρ(x xy) 2 for distinct x, y, z {a, b, c, d} for distinct x, y {a, b, c, d} From the choice data, a b, c d and ab cd. It is easy to verify that the choice data satisfies Axiom T and Axiom WA**. However, 2 β(ab, cd) β(a, cd) + β(b, cd). The choice data does not satisfy weak additivity of relative probability and hence cannot represented by a two-stage Luce rule with attraction system. References Brady, R. L., and J. Rehbeck (205): Menu-Dependent Stochastic Feasibility, mimeo. University of California San Diego. Debreu, G. (960): Review of R.D. Luce, Individual Choice Behavior: A Theoretical Analysis, American Economic Review, 50(), Echenique, F., K. Saito, and G. Tserenjigmid (204): The Perception-Adjusted Luce Model, mimeo. California Institute of Technology. Gul, F., P. Natenzon, and W. Pesendorfer (204): Random Choice as Behavioral Optimization, Econometrica, 82(5),

28 Huber, J., J. W. Payne, and C. Puto (982): Adding Asymmetrically Dominated Alternatives: Violations of Regularity and the Similarity Hypothesis, Journal of Consumer Research, 9(), Luce, R. D. (959): Individual Choice Behavior: A Theoretical Analysis. New York: Wiley. Manzini, P., and M. Mariotti (204): Stochastic Choice and Consideration Sets, Econometrica, 82(3), Manzini, P., M. Mariotti, and L. Mittone (200): Choosing Monetary Sequences: Theory and Experimental Evidence, Theory and Decision, 69(3), McFadden, D. (200): Economic Choices, American Economic Review, 9(3), Rieskamp, J., J. R. Busemeyer, and B. A. Mellers (2006): Extending the Bounds of Rationality: Evidence and Theories of Preferential Choice, Journal of Economic Literature, 44(3), Simonson, I. (989): Choice Based on Reasons: The Case of Attraction and Compromise Effects, Journal of Consumer Research, 6(2), Simonson, I., and A. Tversky (992): Choice in Context: Tradeoff Contrast and Extremeness Aversion, Journal of Marketing Research, 29(3), Sippel, R. (997): An Experiment on the Pure Theory of Consumer s Behaviour, Economic Journal, 07(444), Tversky, A. (969): Intransitivity of Preferences, Psychological Review, 76(), (972a): Choice by Elimination, Journal of Mathematical Psychology, 9(4), (972b): Elimination by Aspects: A Theory of Choice, Psychological Review, 79(4), 28

The Perception-Adjusted Luce Model

The Perception-Adjusted Luce Model The Perception-Adjusted Luce Model Federico Echenique Kota Saito Gerelt Tserenjigmid California Institute of Technology October 1, 2013 Abstract We develop an axiomatic model that builds on Luce s (1959)

More information

Menu-Dependent Stochastic Feasibility

Menu-Dependent Stochastic Feasibility Menu-Dependent Stochastic Feasibility Richard L. Brady and John Rehbeck Last Updated: March 18, 2016 Abstract We examine the role of stochastic feasibility in consumer choice using a random conditional

More information

The Perception-Adjusted Luce Model

The Perception-Adjusted Luce Model The Perception-Adjusted Luce Model Federico Echenique Kota Saito Gerelt Tserenjigmid California Institute of Technology December 22, 2014 Abstract We develop an axiomatic theory of random choice that builds

More information

The Simple Theory of Temptation and Self-Control

The Simple Theory of Temptation and Self-Control The Simple Theory of Temptation and Self-Control Faruk Gul and Wolfgang Pesendorfer Princeton University January 2006 Abstract We analyze a two period model of temptation in a finite choice setting. We

More information

Choice, Consideration Sets and Attribute Filters

Choice, Consideration Sets and Attribute Filters Choice, Consideration Sets and Attribute Filters Mert Kimya Koç University, Department of Economics, Rumeli Feneri Yolu, Sarıyer 34450, Istanbul, Turkey December 28, 2017 Abstract It is well known that

More information

Revealed Reversals of Preferences

Revealed Reversals of Preferences Revealed Reversals of Preferences Houy Nicolas October 5, 2009 Abstract We weaken the implicit assumption of rational choice theory that imposes that preferences do not depend on the choice set. We concentrate

More information

Probabilistic Subjective Expected Utility. Pavlo R. Blavatskyy

Probabilistic 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 information

Great Expectations. Part I: On the Customizability of Generalized Expected Utility*

Great 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 information

UNIVERSITY OF NOTTINGHAM. Discussion Papers in Economics CONSISTENT FIRM CHOICE AND THE THEORY OF SUPPLY

UNIVERSITY OF NOTTINGHAM. Discussion Papers in Economics CONSISTENT FIRM CHOICE AND THE THEORY OF SUPPLY UNIVERSITY OF NOTTINGHAM Discussion Papers in Economics Discussion Paper No. 0/06 CONSISTENT FIRM CHOICE AND THE THEORY OF SUPPLY by Indraneel Dasgupta July 00 DP 0/06 ISSN 1360-438 UNIVERSITY OF NOTTINGHAM

More information

A Random Attention Model

A Random Attention Model A Random Attention Model Matias D. Cattaneo Xinwei Ma U. Michigan U. Michigan Yusufcan Masatlioglu Elchin Suleymanov U. Maryland U. Michigan Stochastic Choice Monotonic Attention RAM Identification Inference

More information

Partially Dominant Choice

Partially Dominant Choice Partially Dominant Choice Georgios Gerasimou February 19, 2015 Abstract This paper proposes and analyzes a model of context-dependent choice with stable but incomplete preferences that is based on the

More information

Regular Choice and the Weak Axiom of Stochastic Revealed Preference

Regular Choice and the Weak Axiom of Stochastic Revealed Preference Regular Choice and the Weak xiom of Stochastic Revealed Preference Indraneel Dasgupta School of Economics, University of Nottingham, UK. and Prasanta K. Pattanaik Department of Economics, University of

More information

Choice, Consideration Sets and Attribute Filters

Choice, Consideration Sets and Attribute Filters Choice, Consideration Sets and Attribute Filters Mert Kimya Brown University, Department of Economics, 64 Waterman St, Providence RI 02912 USA. October 6, 2015 Abstract It is well known that decision makers

More information

A Random Attention Model

A Random Attention Model A Random Attention Model Matias D. Cattaneo Xinwei Ma Yusufcan Masatlioglu Elchin Suleymanov April 2, 2018 Abstract We introduce a Random Attention Model (RAM) allowing for a large class of stochastic

More information

DEPARTMENT OF ECONOMICS DISCUSSION PAPER SERIES

DEPARTMENT OF ECONOMICS DISCUSSION PAPER SERIES ISSN 1471-0498 DEPARTMENT OF ECONOMICS DISCUSSION PAPER SERIES TESTING FOR PRODUCTION WITH COMPLEMENTARITIES Pawel Dziewulski and John K.-H. Quah Number 722 September 2014 Manor Road Building, Manor Road,

More information

Strategy-proof and fair assignment is wasteful

Strategy-proof and fair assignment is wasteful Strategy-proof and fair assignment is wasteful Giorgio Martini June 3, 2016 I prove there exists no assignment mechanism that is strategy-proof, non-wasteful and satisfies equal treatment of equals. When

More information

Rationalization of Collective Choice Functions by Games with Perfect Information. Yongsheng Xu

Rationalization of Collective Choice Functions by Games with Perfect Information. Yongsheng Xu Rationalization of Collective Choice Functions by Games with Perfect Information by Yongsheng Xu Department of Economics, Andrew Young School of Policy Studies Georgia State University, Atlanta, GA 30303

More information

A Rothschild-Stiglitz approach to Bayesian persuasion

A 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 information

Lexicographic Choice under Variable Capacity Constraints

Lexicographic Choice under Variable Capacity Constraints Lexicographic Choice under Variable Capacity Constraints Battal Doğan Serhat Doğan Kemal Yıldız May 14, 2017 Abstract In several matching markets, in order to achieve diversity, agents priorities are allowed

More information

Preferences and Utility

Preferences and Utility Preferences and Utility This Version: October 6, 2009 First Version: October, 2008. These lectures examine the preferences of a single agent. In Section 1 we analyse how the agent chooses among a number

More information

Revealed Attention. Y. Masatlioglu D. Nakajima E. Ozbay. Jan 2011

Revealed Attention. Y. Masatlioglu D. Nakajima E. Ozbay. Jan 2011 Revealed Attention Y. Masatlioglu D. Nakajima E. Ozbay Jan 2011 Limited Attention People do not pay attention to all products because Information Overload, Unawareness, Or both. Breakfast Cereal Breakfast

More information

Comment on The Veil of Public Ignorance

Comment 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 information

This corresponds to a within-subject experiment: see same subject make choices from different menus.

This 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 information

A Rothschild-Stiglitz approach to Bayesian persuasion

A Rothschild-Stiglitz approach to Bayesian persuasion A Rothschild-Stiglitz approach to Bayesian persuasion Matthew Gentzkow and Emir Kamenica Stanford University and University of Chicago September 2015 Abstract Rothschild and Stiglitz (1970) introduce a

More information

A Rothschild-Stiglitz approach to Bayesian persuasion

A 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 information

2 Equivalence Relations

2 Equivalence Relations 2 Equivalence Relations In mathematics, we often investigate relationships between certain objects (numbers, functions, sets, figures, etc.). If an element a of a set A is related to an element b of a

More information

Introductory notes on stochastic rationality. 1 Stochastic choice and stochastic rationality

Introductory notes on stochastic rationality. 1 Stochastic choice and stochastic rationality Division of the Humanities and Social Sciences Introductory notes on stochastic rationality KC Border Fall 2007 1 Stochastic choice and stochastic rationality In the standard theory of rational choice

More information

Non-deteriorating Choice Without Full Transitivity

Non-deteriorating Choice Without Full Transitivity Analyse & Kritik 29/2007 ( c Lucius & Lucius, Stuttgart) p. 163 187 Walter Bossert/Kotaro Suzumura Non-deteriorating Choice Without Full Transitivity Abstract: Although the theory of greatest-element rationalizability

More information

OUTER MEASURE AND UTILITY

OUTER MEASURE AND UTILITY ECOLE POLYTECHNIQUE CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE OUTER MEASURE AND UTILITY Mark VOORNEVELD Jörgen W. WEIBULL October 2008 Cahier n 2008-28 DEPARTEMENT D'ECONOMIE Route de Saclay 91128 PALAISEAU

More information

arxiv: v1 [cs.gt] 10 Apr 2018

arxiv: v1 [cs.gt] 10 Apr 2018 Individual and Group Stability in Neutral Restrictions of Hedonic Games Warut Suksompong Department of Computer Science, Stanford University 353 Serra Mall, Stanford, CA 94305, USA warut@cs.stanford.edu

More information

Random Choice and Learning

Random Choice and Learning Random Choice and Learning Paulo Natenzon First version: November 2010; Updated: April 2015 Abstract The attraction effect and other decoy effects are often understood as anomalies and modeled as departures

More information

Rational Contextual Choices under Imperfect Perception of Attributes

Rational Contextual Choices under Imperfect Perception of Attributes Rational Contextual Choices under Imperfect Perception of Attributes Junnan He Washington University in St. Louis January, 018 Abstract Classical rational choice theory proposes that preferences do not

More information

Characterizing Ideal Weighted Threshold Secret Sharing

Characterizing Ideal Weighted Threshold Secret Sharing Characterizing Ideal Weighted Threshold Secret Sharing Amos Beimel Tamir Tassa Enav Weinreb August 12, 2004 Abstract Weighted threshold secret sharing was introduced by Shamir in his seminal work on secret

More information

Individual decision-making under certainty

Individual decision-making under certainty Individual decision-making under certainty Objects of inquiry Our study begins with individual decision-making under certainty Items of interest include: Feasible set Objective function (Feasible set R)

More information

RANDOM ASSIGNMENT OF MULTIPLE INDIVISIBLE OBJECTS

RANDOM ASSIGNMENT OF MULTIPLE INDIVISIBLE OBJECTS RANDOM ASSIGNMENT OF MULTIPLE INDIVISIBLE OBJECTS FUHITO KOJIMA Department of Economics, Harvard University Littauer Center, Harvard University, 1875 Cambridge St, Cambridge, MA 02138 kojima@fas.harvard.edu,

More information

Numerical representations of binary relations with thresholds: A brief survey 1

Numerical representations of binary relations with thresholds: A brief survey 1 Numerical representations of binary relations with thresholds: A brief survey 1 Fuad Aleskerov Denis Bouyssou Bernard Monjardet 11 July 2006, Revised 8 January 2007 Typos corrected 1 March 2008 Additional

More information

The Revealed Preference Implications of Reference Dependent Preferences

The Revealed Preference Implications of Reference Dependent Preferences The Revealed Preference Implications of Reference Dependent Preferences Faruk Gul and Wolfgang Pesendorfer Princeton University December 2006 Abstract Köszegi and Rabin (2005) propose a theory of reference

More information

Agreeing to disagree: The non-probabilistic case

Agreeing to disagree: The non-probabilistic case Games and Economic Behavior 69 (2010) 169 174 www.elsevier.com/locate/geb Agreeing to disagree: The non-probabilistic case Dov Samet Tel Aviv University, Faculty of Management, Tel Aviv, Israel Received

More information

Asymmetric Dominance, Deferral and Status Quo Bias in a Theory of Choice with Incomplete Preferences

Asymmetric Dominance, Deferral and Status Quo Bias in a Theory of Choice with Incomplete Preferences MPRA Munich Personal RePEc Archive Asymmetric Dominance, Deferral and Status Quo Bias in a Theory of Choice with Incomplete Preferences Georgios Gerasimou 7. July 2012 Online at http://mpra.ub.uni-muenchen.de/40097/

More information

Exercise 1.2. Suppose R, Q are two binary relations on X. Prove that, given our notation, the following are equivalent:

Exercise 1.2. Suppose R, Q are two binary relations on X. Prove that, given our notation, the following are equivalent: 1 Binary relations Definition 1.1. R X Y is a binary relation from X to Y. We write xry if (x, y) R and not xry if (x, y) / R. When X = Y and R X X, we write R is a binary relation on X. Exercise 1.2.

More information

1 The Real Number System

1 The Real Number System 1 The Real Number System The rational numbers are beautiful, but are not big enough for various purposes, and the set R of real numbers was constructed in the late nineteenth century, as a kind of an envelope

More information

A Random Attention Model

A Random Attention Model A Random Attention Model Matias D. Cattaneo Xinwei Ma Yusufcan Masatlioglu Elchin Suleymanov May 30, 2018 Abstract We introduce a Random Attention Model (RAM) allowing for a large class of stochastic consideration

More information

Choosing the two finalists

Choosing the two finalists Choosing the two finalists Economic Theory ISSN 0938-2259 Volume 46 Number 2 Econ Theory (2010) 46:211-219 DOI 10.1007/ s00199-009-0516-3 1 23 Your article is protected by copyright and all rights are

More information

Thus, X is connected by Problem 4. Case 3: X = (a, b]. This case is analogous to Case 2. Case 4: X = (a, b). Choose ε < b a

Thus, X is connected by Problem 4. Case 3: X = (a, b]. This case is analogous to Case 2. Case 4: X = (a, b). Choose ε < b a Solutions to Homework #6 1. Complete the proof of the backwards direction of Theorem 12.2 from class (which asserts the any interval in R is connected). Solution: Let X R be a closed interval. Case 1:

More information

Question 1. (p p) (x(p, w ) x(p, w)) 0. with strict inequality if x(p, w) x(p, w ).

Question 1. (p p) (x(p, w ) x(p, w)) 0. with strict inequality if x(p, w) x(p, w ). University of California, Davis Date: August 24, 2017 Department of Economics Time: 5 hours Microeconomics Reading Time: 20 minutes PRELIMINARY EXAMINATION FOR THE Ph.D. DEGREE Please answer any three

More information

Technical Results on Regular Preferences and Demand

Technical Results on Regular Preferences and Demand Division of the Humanities and Social Sciences Technical Results on Regular Preferences and Demand KC Border Revised Fall 2011; Winter 2017 Preferences For the purposes of this note, a preference relation

More information

NOTES ON COOPERATIVE GAME THEORY AND THE CORE. 1. Introduction

NOTES ON COOPERATIVE GAME THEORY AND THE CORE. 1. Introduction NOTES ON COOPERATIVE GAME THEORY AND THE CORE SARA FROEHLICH 1. Introduction Cooperative game theory is fundamentally different from the types of games we have studied so far, which we will now refer to

More information

Costly Contemplation

Costly Contemplation Costly Contemplation Haluk Ergin MIT December 2003 Abstract We study preferences over opportunity sets. Such preferences are monotone if every opportunity set is at least as good as its subsets. We prove

More information

Preference, Choice and Utility

Preference, 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 information

Certainty Equivalent Representation of Binary Gambles That Are Decomposed into Risky and Sure Parts

Certainty Equivalent Representation of Binary Gambles That Are Decomposed into Risky and Sure Parts Review of Economics & Finance Submitted on 28/Nov./2011 Article ID: 1923-7529-2012-02-65-11 Yutaka Matsushita Certainty Equivalent Representation of Binary Gambles That Are Decomposed into Risky and Sure

More information

UNIVERSITY OF NOTTINGHAM. Discussion Papers in Economics. Contraction consistent stochastic choice correspondence

UNIVERSITY OF NOTTINGHAM. Discussion Papers in Economics. Contraction consistent stochastic choice correspondence UNIVERSITY OF NOTTINGHM Discussion Papers in Economics Discussion Paper No. 08/04 Contraction consistent stochastic choice correspondence y Indraneel Dasgupta June 2008 2008 DP 08/04 Contraction consistent

More information

Assortment Optimization under Variants of the Nested Logit Model

Assortment Optimization under Variants of the Nested Logit Model Assortment Optimization under Variants of the Nested Logit Model James M. Davis School of Operations Research and Information Engineering, Cornell University, Ithaca, New York 14853, USA jmd388@cornell.edu

More information

Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets

Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets Assortment Optimization under the Multinomial Logit Model with Nested Consideration Sets Jacob Feldman School of Operations Research and Information Engineering, Cornell University, Ithaca, New York 14853,

More information

A Note on Object Allocation under Lexicographic Preferences

A Note on Object Allocation under Lexicographic Preferences A Note on Object Allocation under Lexicographic Preferences Daniela Saban and Jay Sethuraman March 7, 2014 Abstract We consider the problem of allocating m objects to n agents. Each agent has unit demand,

More information

Cowles Foundation for Research in Economics at Yale University

Cowles Foundation for Research in Economics at Yale University Cowles Foundation for Research in Economics at Yale University Cowles Foundation Discussion Paper No. 1904 Afriat from MaxMin John D. Geanakoplos August 2013 An author index to the working papers in the

More information

No-envy in Queueing Problems

No-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

13 Social choice B = 2 X X. is the collection of all binary relations on X. R = { X X : is complete and transitive}

13 Social choice B = 2 X X. is the collection of all binary relations on X. R = { X X : is complete and transitive} 13 Social choice So far, all of our models involved a single decision maker. An important, perhaps the important, question for economics is whether the desires and wants of various agents can be rationally

More information

Rationality and solutions to nonconvex bargaining problems: rationalizability and Nash solutions 1

Rationality and solutions to nonconvex bargaining problems: rationalizability and Nash solutions 1 Rationality and solutions to nonconvex bargaining problems: rationalizability and Nash solutions 1 Yongsheng Xu Department of Economics Andrew Young School of Policy Studies Georgia State University, Atlanta,

More information

Dominance and Admissibility without Priors

Dominance 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 information

Topological properties

Topological properties CHAPTER 4 Topological properties 1. Connectedness Definitions and examples Basic properties Connected components Connected versus path connected, again 2. Compactness Definition and first examples Topological

More information

ONLINE APPENDIX TO HOW TO CONTROL CONTROLLED SCHOOL CHOICE (NOT FOR PUBLICATION)

ONLINE APPENDIX TO HOW TO CONTROL CONTROLLED SCHOOL CHOICE (NOT FOR PUBLICATION) ONLINE APPENDIX TO HOW TO CONTROL CONTROLLED SCHOOL CHOICE (NOT FOR PUBLICATION) FEDERICO ECHENIQUE AND M. BUMIN YENMEZ Appendix B. Proofs of Theorems 1-6 The following rationality axiom simply says that

More information

Construction Modelling of Input-Output Coefficients Matrices in an Axiomatic Context: Some Further Considerations.

Construction Modelling of Input-Output Coefficients Matrices in an Axiomatic Context: Some Further Considerations. 1 Construction Modelling of Input-Output Coefficients Matrices in an Axiomatic Context: Some Further Considerations. José Manuel Rueda-Cantuche Universidad Pablo de Olavide Seville - Spain XIVth International

More information

Persuading a Pessimist

Persuading a Pessimist Persuading a Pessimist Afshin Nikzad PRELIMINARY DRAFT Abstract While in practice most persuasion mechanisms have a simple structure, optimal signals in the Bayesian persuasion framework may not be so.

More information

Prudential Choice. January 10, Abstract

Prudential Choice. January 10, Abstract Prudential Choice Serhat Dogan Kemal Yildiz January 10, 2018 Abstract We present a new choice model. An agent is endowed with two sets of preferences: pro-preferences and con-preferences. For each choice

More information

Random Choice and Learning

Random Choice and Learning Random Choice and Learning Paulo Natenzon First version: November 2010; Updated: August 2014 Abstract The attraction effect and other decoy effects are often understood as anomalies and modeled as departures

More information

Identifying Groups in a Boolean Algebra

Identifying Groups in a Boolean Algebra Identifying Groups in a Boolean Algebra Wonki Jo Cho Biung-Ghi Ju June 27, 2018 Abstract We study the problem of determining memberships to the groups in a Boolean algebra. The Boolean algebra is composed

More information

Norm-Constrained Choices

Norm-Constrained Choices Analyse & Kritik 29/2007 ( c Lucius & Lucius, Stuttgart) p. 329 339 Yongsheng Xu Norm-Constrained Choices Abstract: This paper develops a general and unified framework to discuss individual choice behaviors

More information

In N we can do addition, but in order to do subtraction we need to extend N to the integers

In N we can do addition, but in order to do subtraction we need to extend N to the integers Chapter The Real Numbers.. Some Preliminaries Discussion: The Irrationality of 2. We begin with the natural numbers N = {, 2, 3, }. In N we can do addition, but in order to do subtraction we need to extend

More information

Characterizing Ideal Weighted Threshold Secret Sharing

Characterizing Ideal Weighted Threshold Secret Sharing Characterizing Ideal Weighted Threshold Secret Sharing Amos Beimel Tamir Tassa Enav Weinreb October 2, 2006 Abstract Weighted threshold secret sharing was introduced by Shamir in his seminal work on secret

More information

Revisiting independence and stochastic dominance for compound lotteries

Revisiting independence and stochastic dominance for compound lotteries Revisiting independence and stochastic dominance for compound lotteries Alexander Zimper Working Paper Number 97 Department of Economics and Econometrics, University of Johannesburg Revisiting independence

More information

Generalized Hypothesis Testing and Maximizing the Success Probability in Financial Markets

Generalized Hypothesis Testing and Maximizing the Success Probability in Financial Markets Generalized Hypothesis Testing and Maximizing the Success Probability in Financial Markets Tim Leung 1, Qingshuo Song 2, and Jie Yang 3 1 Columbia University, New York, USA; leung@ieor.columbia.edu 2 City

More information

arxiv: v1 [math.pr] 10 May 2014

arxiv: v1 [math.pr] 10 May 2014 On the construction and the cardinality of finite σ-fields arxiv:1405.2401v1 [math.pr] 10 May 2014 P. Vellaisamy, S. Ghosh, and M. Sreehari Abstract. In this note, we first discuss some properties of generated

More information

In N we can do addition, but in order to do subtraction we need to extend N to the integers

In N we can do addition, but in order to do subtraction we need to extend N to the integers Chapter 1 The Real Numbers 1.1. Some Preliminaries Discussion: The Irrationality of 2. We begin with the natural numbers N = {1, 2, 3, }. In N we can do addition, but in order to do subtraction we need

More information

Best Guaranteed Result Principle and Decision Making in Operations with Stochastic Factors and Uncertainty

Best Guaranteed Result Principle and Decision Making in Operations with Stochastic Factors and Uncertainty Stochastics and uncertainty underlie all the processes of the Universe. N.N.Moiseev Best Guaranteed Result Principle and Decision Making in Operations with Stochastic Factors and Uncertainty by Iouldouz

More information

Notes on Supermodularity and Increasing Differences. in Expected Utility

Notes 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 information

Proclaiming Dictators and Juntas or Testing Boolean Formulae

Proclaiming Dictators and Juntas or Testing Boolean Formulae Proclaiming Dictators and Juntas or Testing Boolean Formulae Michal Parnas The Academic College of Tel-Aviv-Yaffo Tel-Aviv, ISRAEL michalp@mta.ac.il Dana Ron Department of EE Systems Tel-Aviv University

More information

Technische Universität Dresden Institute of Numerical Mathematics

Technische Universität Dresden Institute of Numerical Mathematics Technische Universität Dresden Institute of Numerical Mathematics An Improved Flow-based Formulation and Reduction Principles for the Minimum Connectivity Inference Problem Muhammad Abid Dar Andreas Fischer

More information

KIER DISCUSSION PAPER SERIES

KIER DISCUSSION PAPER SERIES KIER DISCUSSION PAPER SERIES KYOTO INSTITUTE OF ECONOMIC RESEARCH Discussion Paper No.992 Intertemporal efficiency does not imply a common price forecast: a leading example Shurojit Chatterji, Atsushi

More information

CONSTRUCTION OF sequence of rational approximations to sets of rational approximating sequences, all with the same tail behaviour Definition 1.

CONSTRUCTION OF sequence of rational approximations to sets of rational approximating sequences, all with the same tail behaviour Definition 1. CONSTRUCTION OF R 1. MOTIVATION We are used to thinking of real numbers as successive approximations. For example, we write π = 3.14159... to mean that π is a real number which, accurate to 5 decimal places,

More information

Deceptive Advertising with Rational Buyers

Deceptive Advertising with Rational Buyers Deceptive Advertising with Rational Buyers September 6, 016 ONLINE APPENDIX In this Appendix we present in full additional results and extensions which are only mentioned in the paper. In the exposition

More information

Title: The existence of equilibrium when excess demand obeys the weak axiom

Title: The existence of equilibrium when excess demand obeys the weak axiom Title: The existence of equilibrium when excess demand obeys the weak axiom Abstract: This paper gives a non-fixed point theoretic proof of equilibrium existence when the excess demand function of an exchange

More information

THREE BRIEF PROOFS OF ARROW S IMPOSSIBILITY THEOREM JOHN GEANAKOPLOS COWLES FOUNDATION PAPER NO. 1116

THREE BRIEF PROOFS OF ARROW S IMPOSSIBILITY THEOREM JOHN GEANAKOPLOS COWLES FOUNDATION PAPER NO. 1116 THREE BRIEF PROOFS OF ARROW S IMPOSSIBILITY THEOREM BY JOHN GEANAKOPLOS COWLES FOUNDATION PAPER NO. 1116 COWLES FOUNDATION FOR RESEARCH IN ECONOMICS YALE UNIVERSITY Box 208281 New Haven, Connecticut 06520-8281

More information

0.Axioms for the Integers 1

0.Axioms for the Integers 1 0.Axioms for the Integers 1 Number theory is the study of the arithmetical properties of the integers. You have been doing arithmetic with integers since you were a young child, but these mathematical

More information

Stochastic Complementarity

Stochastic Complementarity Stochastic Complementarity Paola Manzini Marco Mariotti Levent Ülkü This version: May 2015 Abstract Classical definitions of complementarity are based on cross price elasticities, and so they do not apply,

More information

Controlling versus enabling Online appendix

Controlling versus enabling Online appendix Controlling versus enabling Online appendix Andrei Hagiu and Julian Wright September, 017 Section 1 shows the sense in which Proposition 1 and in Section 4 of the main paper hold in a much more general

More information

Nash Equilibria of Games When Players Preferences Are Quasi-Transitive

Nash Equilibria of Games When Players Preferences Are Quasi-Transitive Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Policy Research Working Paper 7037 Nash Equilibria of Games When Players Preferences

More information

Assortment Optimization under Variants of the Nested Logit Model

Assortment Optimization under Variants of the Nested Logit Model WORKING PAPER SERIES: NO. 2012-2 Assortment Optimization under Variants of the Nested Logit Model James M. Davis, Huseyin Topaloglu, Cornell University Guillermo Gallego, Columbia University 2012 http://www.cprm.columbia.edu

More information

Revealed Preferences and Utility Functions

Revealed Preferences and Utility Functions Revealed Preferences and Utility Functions Lecture 2, 1 September Econ 2100 Fall 2017 Outline 1 Weak Axiom of Revealed Preference 2 Equivalence between Axioms and Rationalizable Choices. 3 An Application:

More information

1 Adeles over Q. 1.1 Absolute values

1 Adeles over Q. 1.1 Absolute values 1 Adeles over Q 1.1 Absolute values Definition 1.1.1 (Absolute value) An absolute value on a field F is a nonnegative real valued function on F which satisfies the conditions: (i) x = 0 if and only if

More information

On path independent stochastic choice.

On path independent stochastic choice. On path independent stochastic choice. David S. Ahn UC Berkeley Federico Echenique Caltech March 30, 2016 Kota Saito Caltech Abstract We investigate the consequences of path independence in stochastic

More information

Existence, Incentive Compatibility and Efficiency of the Rational Expectations Equilibrium

Existence, Incentive Compatibility and Efficiency of the Rational Expectations Equilibrium Existence, ncentive Compatibility and Efficiency of the Rational Expectations Equilibrium Yeneng Sun, Lei Wu and Nicholas C. Yannelis Abstract The rational expectations equilibrium (REE), as introduced

More information

Incomplete Pairwise Comparison Matrices and Weighting Methods

Incomplete Pairwise Comparison Matrices and Weighting Methods Incomplete Pairwise Comparison Matrices and Weighting Methods László Csató Lajos Rónyai arxiv:1507.00461v2 [math.oc] 27 Aug 2015 August 28, 2015 Abstract A special class of preferences, given by a directed

More information

QUASI-PREFERENCE: CHOICE ON PARTIALLY ORDERED SETS. Contents

QUASI-PREFERENCE: CHOICE ON PARTIALLY ORDERED SETS. Contents QUASI-PREFERENCE: CHOICE ON PARTIALLY ORDERED SETS ZEFENG CHEN Abstract. A preference relation is a total order on a finite set and a quasipreference relation is a partial order. This paper first introduces

More information

THE EMPIRICAL IMPLICATIONS OF PRIVACY-AWARE CHOICE

THE EMPIRICAL IMPLICATIONS OF PRIVACY-AWARE CHOICE THE EMPIRICAL IMPLICATIONS OF PRIVACY-AWARE CHOICE RACHEL CUMMINGS, FEDERICO ECHENIQUE, AND ADAM WIERMAN Abstract. This paper initiates the study of the testable implications of choice data in settings

More information

Competition relative to Incentive Functions in Common Agency

Competition relative to Incentive Functions in Common Agency Competition relative to Incentive Functions in Common Agency Seungjin Han May 20, 2011 Abstract In common agency problems, competing principals often incentivize a privately-informed agent s action choice

More information

Information Is Not About Measurability

Information Is Not About Measurability Mathematical Social Sciences 47(2), March 2004, pp. 177-185. Information Is Not About Measurability Juan Dubra Federico Echenique June 25, 2003 Abstract We present a simple example where the use of σ-algebras

More information

Profit Maximization and Supermodular Technology

Profit Maximization and Supermodular Technology Forthcoming, Economic Theory. Profit Maximization and Supermodular Technology Christopher P. Chambers and Federico Echenique January 31, 2008 Abstract A dataset is a list of observed factor inputs and

More information

Boundary Behavior of Excess Demand Functions without the Strong Monotonicity Assumption

Boundary Behavior of Excess Demand Functions without the Strong Monotonicity Assumption Boundary Behavior of Excess Demand Functions without the Strong Monotonicity Assumption Chiaki Hara April 5, 2004 Abstract We give a theorem on the existence of an equilibrium price vector for an excess

More information

Ordered categories and additive conjoint measurement on connected sets

Ordered categories and additive conjoint measurement on connected sets Ordered categories and additive conjoint measurement on connected sets D. Bouyssou a, T. Marchant b a CNRS - LAMSADE, Université Paris Dauphine, F-75775 Paris Cedex 16, France b Ghent University, Dunantlaan

More information