Welfare-maximizing scoring rules VERY PRELIMINARY DRAFT

Size: px
Start display at page:

Download "Welfare-maximizing scoring rules VERY PRELIMINARY DRAFT"

Transcription

1 Welfare-maximizing scoring rules VERY PRELIMINARY DRAFT Adam Giles and Peter Postl 27 February 2012 Abstract We study a cardinal model of voting with three alternatives and three agents whose vnm utilities are private information. We restrict ourselves to voting protocols given by two-parameter scoring rules, as studied by Myerson (2002). We show that a good two-parameter scoring rule reduces welfare losses relative to other, more common, voting rules such as Plurality, Approval voting, and the Borda Count. Oxera, Oxford. Department of Economics, University of Birmingham, Birmingham B15 2TT, United Kingdom, p.postl@bham.ac.uk. 1

2 Any theory of voting which does not allow for intensity of preference is certainly incomplete, and any voting system which does not permit its expression cannot be wholly satisfactory. (from Meek (1975), A transferrable voting system including intensity of preference, Mathematique et science humaines, Vol 50, p ) 1 Introduction Within the large literature on Social Choice, there are two theorems that every economist knows: Arrows impossibility theorem, and the Gibbard- Satterthwaite theorem. These important negative results tell us what we cannot do, not what we can do. In this paper, we look at how far we can get with voting rules in a restricted setting without transferable utility. To do so we use a simple model of voting with three alternatives and three agents who s vnm utilities for the alternatives are private information. Much of the social choice literature proceeds in an axiomatic fashion, imposing desirable properties such as independence of irrelevant alternatives, the Pareto criterion, strategy proofness and so on, whilst dealing with agent s ordinal preferences. By comparison, surprisingly little has been covered in the way of analyzing voting systems by cardinal standards of welfare. To make the problem tractable we restrict attention to a two-parameter family of voting rules known as (A, B)-scoring rules, as studied Myerson (2002). This family of voting rules contains many of the more common voting rules. With out setting we ask: What are equilibrium voting-strategies under (A, B)- scoring rules? After characterizing symmetric equilibrium voting-strategies, we adopt a utilitarian welfare criterion and ask which of the (A, B)-scoring rules generate the highest level ex ante expected utilitarian welfare? The results we present in this paper show that in our simple model, Plurality rule and Negative voting cause high welfare-losses relative to first best. Whilst Approval voting performs much better than either of these rules, it does not perform as well as (A, B)-scoring rules that allow agents to express their relative strength. Also, rules which use agents full ordinal rankings in making a collective choice perform better than those which do not. However, such voting systems do not seem to be particularly widely used. Whilst approval voting does allow for the communication of relative strength of preference, it does not convey unambiguously the agents ordinal preferences. Simple scoring rules (such as the Borda count) suffer the opposite problem, 2

3 in that they make use of agents full ordinal rankings, but neglect information on relative strength of preference. An alternative approach to studying welfare-maximizing voting-rules (and a much more general one) would be to view the problem as one of mechanism design, employing the revelation principle and deriving the optimal incentive compatible direct revelation mechanism. Adopting such an approach in a setting with a continuum of voter-types is very diffi cult. General results do not exist for settings with multi-dimensional private information and nontransferable utility. A simpler approach (also not adopted here for computational reasons) would be to restrict the problem to discrete type spaces, in which case it becomes a linear programming problem. In this case, one could characterize numerically the welfare-maximizing direct mechanism. This approach has been adopted for the case of two agents in a related (but simpler setting) by Börgers and Postl (2009). They find that welfare-losses relative to the first best are very small in their setting. There is a small literature that has studied voting rules and the issue of the communication of intensity of preference. For example Sonnenschein and Jackson (2007) show that in the limit incentive constraints become negligible and the first best can be implemented. Hortala-Vallve (2009) shows that in a case with a finite number of collective decisions the first best cannot be implemented. Our work is related also to that of Apesteguia et al. (2010). They look at which decision rules maximize welfare under various different welfare criteria. While they assume that agents have cardinal utilities, these can only be understood by the agents on an ordinal level. Furthermore, they assume that agents reveal their ordinal rankings truthfully, and find the welfaremaximizing decision rules. In general, they find optimal rules to be some form of scoring rule. For utilitarian welfare, with a uniform distribution of intensity of preference they find that the optimal decision rule corresponds to the Borda count. In our paper, where agents need not vote in line with their true ordinal ranking, we find (A, B)-scoring rules that dominate in welfareterms the Borda count. The remainder of this paper is structured as follows: The model is presented in Section 2. In Section 3 we characterize the equilibrium voting strategies for (A, B)-scoring rules. Section 4 is dedicated to welfare comparisons of the different (A, B)-scoring rules. Section 5 provides concluding remarks. Appendix 1 contains details of numerical results about the welfarelevels associated with different (A, B)-scoring rules. Appendix 2 contains all 3

4 proofs. 2 The Model 2.1 Basic set-up There are n + 1 agents who must collectively choose one alternative from the set {X, Y, Z}. Each agent i I {1,..., n + 1} has a von Neumann Morgenstern utility function u i : {X, Y, Z} R representing his preferences over lotteries on {X, Y, Z}. We denote by u i = (u i X, ui Y, ui Z ) the vector of von Neumann Morgenstern utilities that agent i assigns to outcomes X, Y and Z, respectively. We normalize utilities so that u i [0, 1] 3. These features of the von Neumann Morgenstern utility functions are common knowledge. For i = 1,..., n + 1 we refer to the vector u i as agent i s type. We assume that u i is a random variable whose realization is only observed by agent i. The agents types are stochastically independent, and they are identically distributed. We assume that for each agent i, the von Neumann Morgenstern utilities u i X, ui Y and ui Z are drawn independently from [0, 1] according to distribution function G with strictly positive density g on [0, 1]. 1 The typedistribution is common knowledge. The probability that a voter has any one of the six possible ordinal rankings over the set of alternatives is: ( 1 ( u i X ) ) u i X g(u i Y )du i Y g(u i Z)du i Z g(u i X)du i X = u i Z 2.2 Scoring Rules In the framework of section 2.1, we will study and compare voting mechanisms that are scoring rules as defined in Myerson (2002). Under a scoring rule, each voter is asked to submit a vote vector v i (v i X, vi Y, vi Z ), where vi k denotes the number of points given by agent i to alternative k {X, Y, Z}. 1 As we are dealing with state-dependent expected utility here (with a continuum of states of the world that determine the voters types), each agent s von Neumann Morgentstern utility function is unique up to a positive affi ne transformation with statedependent intercept and state-independent slope (see Ritzberger (2000) for details). We can therefore find, for each agent, an appropriate intercept and slope parameter so that the von Neumann Morgenstern utilities of the three alternatives are in the unit interval for all states of the world. 4

5 The vote vectors of all agents are added up, and the alternative with the most points is chosen. In case of a tie for the most points, an alternative is chosen randomly from among those with the most points, each with equal probability. Different scoring rules are characterized by the different sets of vote vectors that are available to the agents. In this paper we consider a family of scoring rules which are characterized by two parameters A and B such that 0 B A 1. In an (A, B)-scoring rule, each agent must choose a vote vector that is a permutation of either (1, A, 0) or (1, B, 0). That is, the voter must give a maximum of 1 point to one alternative, a minimum of 0 points to some other alternative, and either A or B points to the remaining alternative. Many well-known voting rules are special cases of (A, B)-scoring rules. (A, B)-scoring rules 5

6 For example: 1. Plurality Voting: A = B = 0, so that each agent must choose a vote vector that is a permutation of (1, 0, 0). 2. Borda Count: A = B = 0.5, so that each agent must choose a vote vector that is a permutation of (1, 0.5, 0). 3. Approval Voting: A = 1 and B = 0, so that each agent must choose a vote vector that is a permutation of either (1, 0, 0) or (1, 1, 0). 4. Negative Voting: A = B = 1, so that each agent must choose a vote vector that is a permutation of (1, 1, 0). Figure 1 provides an illustration of all (A, B)-scoring rules. While there is a continuum of (A, B)-scoring rules, the finite number of agents in our set-up implies that (A, B)-scoring rules can be divided into a finite number of equivalence classes. We say that any two scoring rules with parameters (Â, B) and (Ã, B) are equivalent if every pair of corresponding aggregate vote vectors under these two scoring rules generates the same probability distribution over outcomes {X, Y, Z}. 2 Equivalence classes for n + 1 = 3 voters 2 For example, in the case of three agents, if two scoring rules with parameters (Â, B) and (Ã, B) are equivalent, then the two corresponding aggregate vote vectors ( 2 i=1 vi X, 2 i=1 vi Y, 2 i=1 vi Z ) = (2 + B, 1 + 2Â, 0) and ( 2 i=1 ṽi X, 2 i=1 ṽi Y, 2 i=1 ṽi Z ) = (2 + B, 1 + 2Ã, 0) generate the same distribution over {X, Y, Z}. 6

7 For the special case of n+1 = 3 voters, Fig. 2 displays the 70 equivalence classes into which the various (A, B)-scoring rules can be divided. Each black line segment, each black dot, and each white polygon represents a different equivalence class. 3 Equilibrium voting strategies Each scoring rule defines a Bayesian game with n + 1 players/agents. A pure strategy for each agent i is a function that assigns to every type u i of agent i one of the vote vectors available to i under the given scoring rule (we denote by V A,B the set of all vote-vectors available to an agent under a given (A, B)-scoring rule). In this paper, we look for symmetric Bayes Nash Equilibria (BNE) in which the strategy used by all agents displays the following properties: 3 (P1) One point is assigned to highest-ranked alternative, and zero points to the lowest-ranked alternative. More formally, if v i 1 and v i 3 are the scores that agent i assigns to his highest- and lowest-ranked alternatives, resp., then the strategy used by agent i is such that v i 1 = 1 and v i 3 = 0 for all agents i I. (P2) If A > B, the number of points assigned to the middle-ranked alternative depends on the voter s intensity of preference. More formally, let u i 1 denote the vnm utility of agent i s highest-ranked alternative, u i 2 the vnm utility of i s middle-ranked alternative, and u i 3 the vnm utility of i s lowest-ranked alternative. Also, denote by v i 2 the score that agent i assigns to this alternative. The strategy used by agent i is such that: where α [0, 1]. v i 2 = { B if u i 2 < αu i 1 + (1 α)u i 3 A if u i 2 > αu i 1 + (1 α)u i 3 Suppose agent j uses a voting strategy of the type described above. From the point of view of any other agent i j, the strategy adopted by agent j, together with the fact j s type u j is a random variable, implies that j s vote vector v j is a discrete random variable with sample space 3 There may, of course, be other equilibria. However, the level of coordination between agents required to achieve many of these equilibria seems unreasonable. 7

8 V A,B. Given any type u j (consisting of components u j 1 > u j 2 > u j 3), the probability Pr{(1, B, 0) u j 1 > u j 2 > u j 3 } that agent j submits the vote vector v j = (v1, j v2, j v3) j = (1, B, 0) is: ( 1 ( u1 ) ) αu1 +(1 α)u u 3 g(u 2 )du 2 g(u 3 )du 3 g(u 1 )du 1 ( ( u1 ) ) u1 p(α). (1) u 3 g(u 2 )du 2 g(u 3 )du 3 g(u 1 )du and the probability that agent j submits the vote vector v j = (v1, j v2, j v3) j = (1, A, 0) is: Pr{(1, A, 0) u j 1 > u j 2 > u j 3 } = 1 p(α) Lemma 1 For any (A, B)-scoring rule, if the n opponents of agent i adopt the same voting strategy that satisfies property P1 and, if A > B also property P2, then it is a best response for agent i to set v i 1 = 1 and v i 3 = 0. The following proposition states our main characterization result for the case of n = 2 opponents for each agent (i.e. n + 1 = 3 agents in total). Work is ongoing to establish this result for arbitrary n 2. Proposition 1 For every (A, B)-scoring rule with A > B, there exists α (0, 1) such that each agent uses the same voting strategy that satisfies properties P1 and P2. Proposition 1 establishes the existence of a symmetric BNE in voting strategies that satisfy properties P1 and P2. Note that the value α need not be unique. For the purpose of welfare comparisons of (A, B)-scoring rules below, we shall use the value of α that maximizes ex ante expected social welfare whenever there are multiple equilibria. While we have explicitly constructed the proof of Proposition 1 for n+1 = 3 agents, it can be readily generalized to the case where n + 1 is a multiple of 3. If n + 1 is not a multiple of 3, then the equilibrium value α may not be in (0, 1), but may lie in [0, 1] instead. Finally, note that the task of computing analytically fixed point(s) α is very hard. In what follows, we therefore compute numerically the fixed points for specific distribution functions G. 8

9 4 Welfare Comparisons Having characterized symmetric Bayes Nash Equilibria of the game induced by any (A, B)-scoring rule in the presence of n + 1 = 3 agents, we now move on to a comparison of the level of ex ante expected (utilitarian) welfare across all possible (A, B)-scoring rules. In order to have a benchmark, we first compute ex ante expected welfare in the hypothetical first best scenario in which the types u i of all three voters, once realized, become observable to the social planner. 4.1 First best welfare A social planner who aims to maximize ex ante expected social welfare must choose for each profile u (u 1, u 2, u 3 ) of the three agents types the alternative in {X, Y, Z} that maximizes the sum of the agents ex post utilities. More formally, the planner chooses a decision rule f : [0, 1] 3 [0, 1] 3 [0, 1] 3 ({X, Y, Z}), u (f X (u), f Y (u), f Z (u)), where ({X, Y, Z}) is the set of probability distributions over {X, Y, Z}, and f k (u) denotes the probability that alternative k {X, Y, Z} is chosen when the agents type-profile is u. Ex ante expected social welfare associated with a decision rule f is: [ E u k {X,Y,Z} f k(u) ( ) ] i I ui k We say that a decision rule f is first best if for every k {X, Y, Z}: i I ui k > max { i I l {X,Y,Z}\k ui l} f k (u) = 1 i I ui k < max { i I l {X,Y,Z}\k ui l} f k (u) = 0 In order to compute ex ante welfare of a first best decision rule, we define three random variables w k i I ui k (one for each k {X, Y, Z}), which has support [0, 3]. Due to the i.i.d. nature of all vnm utilities u i k, each random variable w k has the same cumulative distribution function H. We can now compute first best welfare as the expectation of the random variable max{w X, w Y, w Z }, which has the following cumulative distribution function: H(w) = Pr(max{w X, w Y, w Z } w) = (H(w)) 3 9

10 4.2 Welfare of (A, B)-scoring rules In order to compute welfare losses (relative to first best welfare) under the different (A, B)-scoring rules, we need to compute expected welfare in the BNE of the various (A, B)-scoring rules. To do this, due to the symmetry of our set-up, all we need to do is to compute the expected utility of one agent, and then multiply it by 3 (as we are considering here the case of n + 1 = 3 agents). Observe that the expected utility of any single agent is: U i (u i ) { Ui (B, α, u i ) if u i 2 < α u i 1 + (1 α )u i 2 U i (A, α, u i ) if u i 2 > α u i 1 + (1 α )u i 2 where U i (σ, α, u i ) (with σ {A, B}) is the expected utility of agent i of type u i when he submits the vote-vector (v1, i v2, i v3) i = (1, σ, 0) that corresponds to his true ordinal ranking and in which σ points are assigned to i s secondranked alternative. 4 Due to the symmetry of our set-up, we can compute the ex ante expected utility E[U i (u i )] of agent i by multiplying by 6 (as there are six possible ordinal rankings of the alternatives in {X, Y, Z}) agent i s expected utility for an arbitrary ordinal ranking (as implied by vnm utilities u i 1 > u i 2 > u i 3, where u i l is agent i s vnm utility from his l-th ranked alternative): ( 1 ( u1 ) ) α u 1 +(1 α )u 3 6 U i (B, α, u i )g(u 2 )du 2 g(u 3 )du 3 g(u 1 )du u 3 1 ( u1 ( u1 ) ) +6 U i (A, α, u i )g(u 2 )du 2 g(u 3 )du 3 g(u 1 )du α u 1 +(1 α )u Bounding second best welfare - INCOMPLETE!! While first best welfare is an important benchmark, it is not entirely satisfactory for measuring the performance of different (A, B)-scoring rules. The reason is that first best decision rules do not take into account the incentive constraints of individual agents, who, when given the opportunity to express directly they vnm utilities to the social planner, may find it in their interest to misrepresent their vnm utilities. A better benchmark for the evaluation 4 Note that the BNE value of α is fully determined by the model primitives, which is why the ex ante expected utility of agent i does not depend on α but only on his type u i. 10

11 of (A, B)-scoring rules would therefore be the optimal decision rule among all decision rules that are incentive compatible (we refer to such decision rules as second best decision rules). Definition 1 A decision rule f is incentive compatible if for all i I and for all u i, û i [0, 1] 3 we have: k {X,Y,Z} E u i[f k(û i, u i )]u i k }{{} =U i (û i,u i ) k {X,Y,Z} E u i[f k(u i, u i )]u i k }{{} =U i (u i,u i ) µ i (u i ) The following lemma states standard results from the mechanism design literature for our set-up: Lemma 2 A DRM f is incentive compatible if and only if: 1. µ i is convex, 2. At every u i at which µ i is differentiable: µ i (ui ) u i k all k {X, Y, Z}. = E u i[f k (u i, u i )] for Convexity of µ i implies that agent i s conditional expected probability of alternative k, E u i[f k (u i, u i )], is increasing in agent i s vnm utility of u i k : E u i[f k (u i, u i )] u i k 0 for all k {X, Y, Z}. Following Hortala-Vallve (2009), we can state that if a decision rule f is incentive compatible, then µ i is homogeneous of degree one: µ i (λu i ) = λµ i (u i ) (for λ R ++ ). Homogeneity of µ i implies that the decision rule f bunches all proportional types. By definition, µ i (λu i ) = k {X,Y,Z} E u i[f k(λu i, u i )] (λu i k ). Then, µ i (λu i ) = λµ i (u i ) E u i[f k (λu i, u i )] = E u i[f k (u i, u i )] Observe that homogeneity of degree zero imposes the following restrictions on incentive compatible decision rules f: Take two types of agent i: (u i X, ui Y, ui Z ) and (û i X, ûi Y, ûi Z ), and suppose X i Y i Z. Under state-dependent expected utility, the following transformation of agent i s vnm utilities can be performed wlog: (u i X ui Z, ui Y ui Z, 0) and (ûi X ûi Z, ûi Y ûi Z, 0), which 11

12 obtains by subtracting the vnm utility of the lowest-ranked alternative from the vnm utilities of the other alternatives. Note that, in general, we cannot wlog divide these transformed types by different constants. However, incentive compatibility (in particular, homogeneity of degree zero) implies that the decision rule f must treat a type (u i X ui Z, ui Y ui Z, 0) the same as the type (1, ui Y ui Z, 0), and must treat a type (û i u i X ui X ûi Z, ûi Y ûi Z, 0) the same as Z the type (1, ûi Y ûi Z, 0). û i X ûi Z While incentive compatibility may impose further constraints on decision rules f beyond homogeneity of degree zero, it is clear that no second best decision rule can generate higher ex ante welfare than a decision rule that respects only the homogeneity requirement. Under such a decision rule, any two agents i and j who have the same ordinal ranking, and for whom u i 2 u i 3 u i 1 u i 3 = uj 2 u j 3 u j 1 u j 3 have the same expected utility, and therefore the same best response function. Therefore, there always exists a BNE of the game induced by the decision rule f such that agents i and j report the same message to the social planner. We now state the following conjecture (as yet unproven) which should provide a more realistic benchmark for the evaluation of the welfare-performance of different (A, B)-scoring rules: Conjecture 1 Among decision rules f for which the bunching of proportional types -condition holds, the optimal decision rule assigns to each agent i who reports his ordinal ranking i over {X, Y, Z}, as well as a number θ i = (u i 2 u i 3)/(u i 1 u i 3), an expected type-vector τ i. In this expected typevector τ i, the alternative that agent i identifies as his highest-ranked alternative is assigned a value of E[u 1 θ i ]; the alternative that he identifies as his middle-ranked alternative is assigned a value of θ i E[u 1 θ i ]+(1 θ i )E[u 3 θ i ]; and the alternative that he identifies as his lowest-ranked alternative is assigned a value of E[u 3 θ i ]. On the basis of expected types τ i for all agents, the decision rule selects the one that would be first best if the τ i were the agents true types. 4.4 Numerical results This section accompanies the diagrams and tables found in Appendix 1. Looking at the diagrams indicating welfare losses, it is clear that (A, B)- 12

13 scoring rules with large A and relatively small B provide the lowest welfare loss. In fact in each case, the best (A, B)-scoring rule is any one from the open set of points marked in grey in the Figure 3 below. Negative voting is consistently the worst rule, indicated in the figures by the black dot at (1,1). The plurality rule is consistently the second-worst rule, indicated in the figures by the dark grey dot at (0,0). The best simple scoring rule (i.e. rules for which A = B) is the one that is consistent with the results in Apesteguia et al. (2010). Approval voting performs better than either negative voting or plurality rule, but not as well as the best simple scoring rule, or the optimal (A, B)- scoring rules shown in Fig. 3. Almost all rules that satisfy 1 > A > B > 0 yield higher welfare than those that do not, which is due to the fact that these rules use the agents full ordinal rankings in making a collective choice. One surprising fact that is evident from Fig. 3. is that it is possible to pick an (A, B)-pair that is extremely close to (1, 0) (approval voting), but yields substantially higher welfare. This discontinuity stems from the discrete number of agents in the model. Our conjecture in the previous section sharpens the intuition for our numerical results. To highlight this, we consider (for simplicity) in the remainder of this section the case where the distribution G from which the agents individual vnm utilities are drawn is the uniform distribution on [0, 1]. In this case, E[u i 1 θ i ] = E[u i 1] = 0.75, and E[u i 3 θ i ] = E[u i 3] = 0.25 for for every θ i and any i I. Therefore, the expected type of agent i in the decision rule described in Conjecture 1 is: τ i = (τ i 1, τ i 2, τ i 3) = (0.75, 0.75θ i (1 θ i ), 0.25) Applying an affi ne transformation where τ i l = (τ i l τ i 3)/(τ i 1 τ i 3) yields: τ i = ( τ i 1, τ i 2, τ i 3) = (1, θ i, 0) Now consider an (A, B)-scoring rule in gray shaded area of Fig. 3. For scoring rules in this equivalence class, the unique value α that characterizes the symmetric BNE voting strategy is α = Now consider the particular (A, B)-scoring rule where A = 0.9 and B = 0.4, which is contained in the shaded area in Fig. 3. Note that any agent i whose type u i gives rise to a number θ i > submits a vote-vector that is a permutation of (1, 0.9, 0), while any type u i with θ i < submits a vote-vector that is a permutation of (1, 0.4, 0). These two vote-vectors are approximately equal 13

14 to the expected types who use these vote-vectors in equilibrium: (1, 0.9, 0) E[u i θ i > α ] and (1, 0.4, 0) E[u i θ i < α ]. 5 In other words, conditional on an agent submitting the vote-vector (1, 0.9, 0), his expected value of θ i is approx. 0.9, and condition al on an agent submitting the vote-vector (1, 0.4, 0), his expected value of θ i is approx Under Approval voting, which performs less well than any rule from the shaded set in Fig. 3., the equilibrium value of α is α = 0, Types with a value of θ i > submit a vote-vector that is a permutation of (1, 1, 0), while types with a value of θ i < submit a vote-vector that is a permutation of (1, 0, 0). Clearly, these vote-vectors do not represent very well the average type of agent who uses them, hence the higher welfare-loss relative to the benchmark decision rule in Conjecture 1. The same can be said of Negative Voting, and the plurality rule. The Borda count, which is the optimal simple scoring rule (i.e. one where A = B), effectively assigns all agents an expected type of τ i = ( τ i 1, τ i 2, τ i 3) = (1, 0.5, 0). This is also equal to the expected type E[u i θ i ] of each agent who uses this vote-vector, but as it assigns it to every type, it is not as effective as the optimal best (A, B)-scoring rule. Numerically calculating expected welfare under the benchmark mechanism in Conjecture 1 using Monte-Carlo experiments for the case of the uniform distribution yields a welfare-loss against the First Best of 1.604%. The welfare loss of the best (A, B)-scoring rule against the benchmark mechanism in Conjecture 1 is, for the uniform case, equal to 0.458%. This value is very small, and suggests that most of the welfare-loss can be attributed to the necessity of bunching all proportional types. This observation fits well with the finding Börgers and Postl (2009), who noted that welfare-losses were almost negligible in their setting, where agents types were normalized so that each agent s most preferred alternative had a vnm utility of 1, and each agent s least preferred alternative had a vnm utility 0. Consequently, there were no proportional types that had to be bunched. Finally, looking at the tables in Appendix 1, it is clear that - although not optimal - the best simple scoring rules often come close to the best (A, B)- scoring rule in terms of expected welfare. 5 The actual expected value of θ i among those who submit the vote-vector (1, 0.9, 0) is , and the actual expected value of θ i among those who submit the vote-vector (1, 0.4, 0) is Note that the pair (A, B) = (0.8963, ) constitutes a twoparameter scoring rule in the equivalence class highlighted in Fig

15 5 Conclusion The aim of this paper was to study cardinal standards of welfare in voting problems without transferable utility using (A, B)-scoring rules. Our results show that in this simple model, Plurality rule and Negative voting cause high welfare-losses relative to first best. Whilst Approval voting performs much better than either of these rules, it does not perform as well as rules with a relatively large A-value and small B-value. More generally, it seems that voting rules that allow agents to express relative strength of preference perform better than those which do not. Also, rules which use agents full ordinal rankings perform better than those which do not. However, such systems do not seem to be particularly widely used. Whilst approval voting does involve relative strength of preference, the resulting vote vectors do not convey unambiguously the agents ordinal preferences. Simple scoring rules (for which A = B) suffer the opposite problem. In future work we intend to extend our model to settings with a larger number of agents or alternatives. With a larger number of agents, one would expect the highly discontinuous nature of the welfare-levels across (A, B)-scoring rules to be smoothed out by making the equivalence classes into which (A, B)-scoring rules can be divide much finer. Another extension might look into optimality of different rules in a common value setting, as opposed to the private value setting covered here. There is some research already in this area, for example Ahn and Oliveros (2010) use (A, B)-Scoring rules to show that in a common value setting with three alternatives, the best equilibria under Approval voting are more effi cient than Plurality of Negative voting. Finally, future research might also look at the continued use of numerical optimization techniques in a discrete setting. Whilst the issue of characterizing optimal direct revelation mechanisms for more than two agents remains an open problem, improvements in computing power should allow the use of finer discrete type-sets than has been possible so far. 6 References Ahn, D., Oliveros, S., Approval voting and scoring rules with commonvalues. UC Berkeley Mimeo. Apesteguia, J., Ballester, M. A., Ferrer, R., decision rules. Review of Economic Studies 78 (1), On the justice of

16 Borgers, T., Postl, P., Effi cient compromising. Journal of Economic Theory 144, Hortala-Vallve, R., Ineffi ciencies on linking decisions. Social Choice and Welfare 34, Myerson, R., Comparison of scoring rules in poisson voting games. Journal of Economic Theory 103 (1), Postl, P., A bayesian mechanism design approach to optimal voting rules when agents preferences are private information. University of Birmingham mimeo. Sonnenschein, H., Jackson, M., Overcoming incentive constraints by linking decisions. Journal of Economic Theory 75 (1), Appendix 1 The following tables and figures provide the relative welfare losses of a number of (A, B)-scoring rules for three distributions G, each given by a different Beta-distribution. The figures represent the ex ante expected welfare in the BNE of every possible (A, B)-scoring rule. In particular, the figures show the relative welfare-loss relative to first best (the welfare of which has been calculated using Monte-Carlo experiments). The lowest welfare-loss is indicated in white, while black indicates the highest welfare-loss. Intermediate welfarelosses are shown in different shades of gray. The three Beta-distributions that we have chosen in this paper are: Beta(1,1) (which corresponds to the uniform distribution), Beta(2,1) (which places more probability mass on high realizations), and Beta (1,2) (which places more probability mass on low realizations). 16

17 17

18 Relative welfare-loss for Beta(1,1) (uniform) Relative welfare-loss for Beta(1,2) 18

19 Relative welfare-loss for Beta(2,1) 8 Appendix 2 Proof. Step 1: Take any agent and a profile of his n opponents vote vectors such that alternative X has the highest score. I.e. the aggregate vote vector v = ( j i vj X, j i vj Y, j i vj Z ) is such that j i vj X > max{ j i vj Y, j i vj Z }. Now, in the individual vote vector of each agent j, interchange the scores assigned to alternatives X and Y : v j = (v j Y, vj X, vj Z ). This gives rise to the aggregate vote vector v = ( j i vj Y, j i vj X, j i vj Z ) under which alternative Y as the highest score. Similarly, by interchanging the scores assigned to alternatives X and Z in the individual vote vectors of the n opponents, we can obtain the aggregate vote vector ṽ = ( j i vj Z, j i vj Y, j i vj X ) such that alternative Z has the highest aggregate score. This shows that, for every aggregate vote vector v VA,B n s.t. alternative X has the highest score, there exist vectors v, ṽ VA,B n s.t. alternative Y (alternative Z, resp.) has the highest aggregate score. Therefore, the number of aggregate vote vectors for which alternative Y has the highest score equals the number of aggregate vote vectors for which alternative X has the highest score. Likewise, the number of aggregate vote vectors for which alternative Z has the highest score equals the number of aggregate vote vectors for which alternative X has the highest score. Step 2: Take any agent and a profile of his n opponents vote vectors such that alternative X has the same aggregate score as alternative Y, but higher than the aggregate score of alternative Z. I.e. the aggregate vote vector v = ( j i vj X, j i vj Y, j i vj Z ) across the n opponents is such that j i vj X = j i vj Y > j i vj Z (i.e. the social ranking of the three alternatives by the n opponents is X n Y n Z, where n denotes the social preference relation of the n opponents over outcomes in {X, Y, Z}). Now, in the individual vote vector of each agent j, interchange the scores assigned to alternatives X and Z: ṽ j = (v j Z, vj Y, vj X ). This gives rise to the aggregate vote vector ṽ = ( j i vj Z, j i vj Y, j i vj X ) under which alternatives Y and Z have the same aggregate score that exceeds the score of alternative X (i.e. the social ranking is Y S Z S X). I.e. for every aggregate vote vector in VA,B n s.t. alternatives X and Y tie while beating alternative Z, there exists an aggregate vote vector in VA,B n s.t. alternatives Y and Z tie while beating 19

20 alternative X. Consequently, the number m of aggregate vote vectors in VA,B n s.t. X and Y tie while beating Z is the same as the number of aggregate vote vectors m in VA,B n s.t. Y and Z tie while beating alternative X. Now consider an aggregate vote vector v in VA,B n s.t. alternatives X and Z tie while while beating alternative Y (social ranking: X n Z n Y ). By interchanging the X- and Y -scores in the individual vote vectors, we obtain an aggregate vote vector s.t. alternatives Y and Z tie while beating alternative X (social ranking: Y n Z n X). I.e. for every aggregate vote vector in VA,B n s.t. alternatives X and Z tie while beating alternative Y, there exists another aggregate vote vector in VA,B n s.t. alternatives Y and Z tie while beating alternative X. Consequently, the number m of aggregate vote vectors in VA,B n s.t. X and Z tie while beating Y is the same as the number of aggregate vote vectors m in VA,B n s.t. B and C tie while beating alternative A. Therefore: m = m = m. Step 3: For each voter j, the probability that a permutation v j of a given vote vector v j arises is the same as the probability that v j arises. E.g. Pr{(1, A, 0)} = 1 (1 p(α)), where p(α) is the probability that, conditional on 6 a given ordinal ranking, voter j assigns the lower score of B to his middleranked alternative. We have Pr{(A, 1, 0)} = 1 (1 p(α)) = Pr{(1, A, 0)}, 6 which is due to the fact that all ordinal rankings are equally likely, and that all voters, conditional on a given ordinal ranking, use the same strategy to decide whether to assign a score of B or a score of A to their middle-ranked alternative. On the basis of Steps 1-3, we can immediately conclude that the probability that alternative X has the highest score across the n opponents of agent i is the same as the probability that alternative Y has the highest score, which, in turn, is the same as the probability that alternative Z has the highest score. I.e. we have Pr{X} = Pr{Y } = Pr{Z} = 1/3. Step 4: Regardless of the value of n 2, there always exist aggregate vote vectors s.t. X n Y n Z, and consequently, there exist equally many aggregate vote vectors s.t. X n Z n Y and s.t. Y n Z n X. In addition, if there are n = 3 opponents (or any multiple of 3), there are also vote vectors s.t. X n Y n Z. However, such three-way ties do not exist for n = 2 opponents, or if n/3 is not an integer (e.g. n = 148 or n = 55. For n = 55, I can create two-way ties even though 55/2 is not an integer. To see this, suppose that three opponents have vote vectors so that there is a three-way tie. E.g. (1, A, 0), (0, 1, A) and (A, 0, 1), which gives an aggregate of (1+A, 1+A, 1+A) across these three opponents. Now let (55 3)/2 = 26 20

21 submit the vote vector (1, B, 0), while the remaining 26 submit (B, 1, 0). This yields an aggregate vote vector across all n = 55 voters of (1 + A + 26(1 + B), 1 + A + 26(B + 1), 1 + A) = ( B + A, B + A, 1 + A)). Step 5: Now consider agent i and suppose he submits a vote vector that assigns a score of 1 to alternative U, a score of σ {A, B} to alternative V, and a score of 0 to alternative W, where U, V, W {X, Y, Z}, U V W. Now consider aggregate vote vectors across the n opponents of agent i where U ties with one of the other two alternatives. If U n V n W, we now have ( j i vj U + 1, j i vj U + σ, j i vj W ) so that U n+1 V n+1 W ; and if U n W n V, now have ( j i vj U + 1, j i vj V + σ, j i vj U ) so that j i vj U + 1 max{ j i vj V + σ, j i vj U }. In each such event where agent i s vote vector generates an outright win for alternative U (and there is a positive number of such events), alternative U gains probability weight 1/2 at the expense of alternatives V and W, resp. As the probability distribution across the three alternatives prior to accounting for voter i s vote vector was uniform (Pr{U} = Pr{V } = Pr{W } = 1/3), it now holds that Pr{U} > 1 > 3 max{pr{v }, Pr{W }}. Obviously, in case that there exist three-ways ties, alternative U gains yet more probability weight at the expense of alternatives V and W. Furthermore, in all events where U narrowly loses to either V or W, alternative U will gain yet more probability weight at the expense of the other two alternatives. Step 6: Now suppose wlog that voter i s ordinal ranking is X i Y i Z, with u i X > ui Y > ui Z. As voter i s expected utility is a weighted sum of these three vnm utilities, it is immediate that as much probability mass should be concentrated on alternative on u i X, and as little as possible on alternative on u i Z. It is therefore optimal for voter i to submit the vote vector that assigns a score of 1 to alternative X, and a score of 0 to alternative Z. Proof. Wlog consider agent i. Suppose agents j i use the same strategy, and that this strategy satisfies P1 and P2, with parameter α [0, 1]. The probability that any agent j i submits the vote vector (v j X, vj Y, vj Z ) = (1, B, 0), or any permutation thereof, is given by 1 p(α), where p(α) is given in 6 (1). The probability that any agent j submits the vote vector (v j X, vj Y, vj Z ) = (1, A, 0), or any permutation thereof, is given by 1 (1 p(α)). Therefore, the 6 aggregate vote vectors in VA,B 2 will occur either with probability ( 1 6 p(α))2, with probability 1p(α)( 1(1 p(α))), or with probability ( (1 p(α)))2. Note that the probability p(α) is a continuous function of α, with p(0) = 0 and p(1) = 1. 21

22 By Lemma 1, agent i s choice of vote vector reduces to the question of whether to set v i 2 = A, or v i 2 = B. Observe that relative to setting v i 2 = B, choosing v i 2 = A (i.e. submitting the vote-vector (v i 1, v i 2, v i 3) = (1, A, 0)) implies a shift in probability mass to alternative 2 from either alternative 1, or alternative 3, or both. To see this, consider the following aggregate vote-vector across all agents j i: ( j i vj 1, j i vj 2, j i vj 3) = (A, 1, 1 + A) This vote vector occurs with probability ( 1 6 (1 p(α)))2, which is positive for α < 1. If agent i submits the vote-vector (v i 1, v i 2, v i 3) = (1, B, 0), the aggregate vote-vector across all three agents is: ( j I vj 1, j I vj 2, j I vj 3) = (1 + A, 1 + B, 1 + A) This implies that alternatives 1 and 2 are both chosen with probability 1/2. If, instead, agent i submits the vote vector (v i 1, v i 2, v i 3) = (1, A, 0), the aggregate vote-vector across all three agents is: ( j I vj 1, j I vj 2, j I vj 3) = (1 + A, 1 + A, 1 + A) This implies that all three alternatives are chosen with equal probability of 1/3. Next, consider the following aggregate vote-vector across all agents j i: ( j i vj 1, j i vj 2, j i vj 3) = (B, 1, 1 + B) This vote vector occurs with probability ( 1 6 p(α))2, which is positive for α > 0. If agent i submits the vote-vector (v i 1, v i 2, v i 3) = (1, B, 0), the aggregate votevector across all three agents is: ( j I vj 1, j I vj 2, j I vj 3) = (1 + B, 1 + B, 1 + B) This implies that all three alternatives are chosen with equal probability of 1/3. If, instead, agent i submits the vote vector (v i 1, v i 2, v i 3) = (1, A, 0), the aggregate vote-vector across all three agents is: ( j I vj 1, j I vj 2, j I vj 3) = (1 + B, 1 + A, 1 + B) This implies that alternative 2 is chosen. The above two examples establish that for any α [0, 1], submitting vote-vector (v i 1, v i 2, v i 3) = (1, A, 0) implies 22

23 (relative to the case where vote vector (v1, i v2, i v3) i = (1, B, 0) is submitted) a strict increase in the probability that alternative 2 is chosen, at the expense of the probabilities with which alternatives 1 and 3 are chosen. Define now by π σ i,l (α) (where l = 1, 2, 3) the probability that agent i s l-th ranked alternative is chosen if i submits the vote vector (v1, i v2, i v3) i = (1, σ, 0) (where σ {A, B}), and the other two agents j i use the same strategy satisfying P1 and P2. The probability π σ i,l (α) is of the following form: π σ i,l(α) = a σ ( 1 6 p(α))2 + b σ 1 6 p(α)(1 6 (1 p(α))) + cσ ( 1 6 (1 p(α)))2 (2) where a σ denotes the ex post probability that alternative 1 is chosen when agent i submits the vote-vector (v1, i v2, i v3) i = (1, σ, 0) and the vote-vectors submitted by the other two agents j i are both permutations of (1, B, 0). Note that, given our tie-breaking rule, the ex post probability of an alternative being chosen is either 1 (if this alternative has a higher aggregate score than the other two alternatives), or 1/2 (if the aggregate score of this alternative equals the aggregate score of one other alternative, and this score exceeds that of the remaining alternative), or 1/3 (if the aggregate scores of all three alternatives are the same). Likewise, b σ denotes the ex post probability that alternative 1 is chosen when agent i submits the vote-vector (v1, i v2, i v3) i = (1, σ, 0), while one other agent submits a vote-vector that is a permutation of (1, B, 0), and the remaining agent submits a vote-vector that is a permutation of (1, A, 0). Finally, c σ denotes the ex post probability that alternative 1 is chosen when agent i submits the vote-vector (v1, i v2, i v3) i = (1, σ, 0), while the vote-vectors submitted by the other two agents j i are both permutations of (1, A, 0). We can then write agent i s expected utility from submitting either of his two vote vectors (1, A, 0) and (1, B, 0) as follows: U i (A, α, u i ) π A i,1(α)u 1 + π A i,2(α)u 2 + π A i,3(α)u 3 U i (B, α, u i ) π B i,1(α)u 1 + π B i,2(α)u 2 + π B i,3(α)u 3 The argument in the second and third paragraphs of this proof implies that: π A i,1(α) < π B i,1(α) π A i,2(α) > π B i,2(α) π A i,3(α) < π B i,3(α) 23

24 As π A i,1(α) + π A i,2(α) + π A i,3(α) = 1 = π B i,1(α) + π B i,2(α) + π B i,3(α), it is immediate that: π A i,2(α) π B i,2(α) = [ π B i,1(α) π A i,1(α) ] + [ π B i,3(α) π A i,3(α) ] (3) As the increase π A i,2(α) π B i,2(α) in the probability of alternative 2 under vote-vector (1, A, 0), relative to vote-vector (1, B, 0), is obtained by transferring probability from alternative 1 and/or alternative 3 to alternative 2, it follows that the left-hand side of (3) is strictly greater than either of the two bracketed terms on the right-hand side of (3). We can re-write (3) as follows: 1 = πb i,1(α) π A i,1(α) π A i,2 (α) πb i,2 (α) + πb i,3(α) π A i,3(α) π A i,2 (α) πb i,2 (α) π B i,3(α) π A i,3(α) π A i,2 (α) πb i,2 (α) = 1 πb i,1(α) π A i,1(α) π A i,2 (α) πb i,2 (α) (4) Now observe that agent i will submit vote-vector (1, A, 0) (rather than (1, B, 0)) iff: U i (A, α, u i ) U i (B, α, u i ) > 0 [ π A i,2 (α) π B i,2(α) ] u i 2 > [ π B i,1(α) π A i,1(α) ] u i 1 + [ π B i,3(α) π A i,3(α) ] u i 2 Using (4), we can re-write this inequality as follows: ( ) u i 2 > πb i,1(α) π A i,1(α) π A i,2 (α) πb i,2 (α)ui πb i,1(α) π A i,1(α) π A i,2 (α) πb i,2 (α) Given the expression for the probabilities π σ i,l in (2), it holds that the coeffi cient [π B i,1(α) π A i,1(α)]/[π A i,2(α) π B i,2(α)] of u i 1 is in (0, 1) for every α [0, 1]. 6 Furthermore, the coeffi cient [π B i,1(α) π A i,1(α)]/[π A i,2(α) π B i,2(α)] is a continuous function of α. In order to obtain a symmetric BNE in which 6 We have verified this by computing for all 0 B < A 1 the chosen alternative(s) in the cases where: (i) the two agents j i submit vote-vectors that are both permutations of (1, B, 0), (ii) the two agents j i submit vote-vectors that are both permutations of (1, A, 0), and (iii) one agent submits vote-vectors that are permutations of (1, B, 0) and the other submits vote-vectors that are permutations of (1, A, 0). u i 3 24

25 agent i adopts the same voting strategy as the other two agents j i (and which satisfies P1 and P2, with some α [0, 1]), it must hold that: π B i,1(α) π A i,1(α) π A i,2 (α) πb i,2 (α) = α By appealing to Brouwer s Fixed Point Theorem, we can conclude that the function [π B i,1(α) π A i,1(α)]/[π A i,2(α) π B i,2(α)] has at least one fixed point α, and all fixed points lie in the open unit interval (0, 1). 25

"Arrow s Theorem and the Gibbard-Satterthwaite Theorem: A Unified Approach", by Phillip Reny. Economic Letters (70) (2001),

Arrow s Theorem and the Gibbard-Satterthwaite Theorem: A Unified Approach, by Phillip Reny. Economic Letters (70) (2001), February 25, 2015 "Arrow s Theorem and the Gibbard-Satterthwaite Theorem: A Unified Approach", by Phillip Reny. Economic Letters (70) (2001), 99-105. Also recommended: M. A. Satterthwaite, "Strategy-Proof

More information

Political Economy of Institutions and Development. Lectures 2 and 3: Static Voting Models

Political Economy of Institutions and Development. Lectures 2 and 3: Static Voting Models 14.773 Political Economy of Institutions and Development. Lectures 2 and 3: Static Voting Models Daron Acemoglu MIT February 7 and 12, 2013. Daron Acemoglu (MIT) Political Economy Lectures 2 and 3 February

More information

FAIRNESS IN MECHANISM DESIGN. February 13, 2015

FAIRNESS IN MECHANISM DESIGN. February 13, 2015 FAIRNESS IN MECHANISM DESIGN YARON AZRIELI AND RITESH JAIN February 13, 2015 Abstract. In a standard Bayesian environment with independent private values and two possible alternatives, it is shown that

More information

Algorithmic Game Theory Introduction to Mechanism Design

Algorithmic Game Theory Introduction to Mechanism Design Algorithmic Game Theory Introduction to Mechanism Design Makis Arsenis National Technical University of Athens April 216 Makis Arsenis (NTUA) AGT April 216 1 / 41 Outline 1 Social Choice Social Choice

More information

APPLIED MECHANISM DESIGN FOR SOCIAL GOOD

APPLIED MECHANISM DESIGN FOR SOCIAL GOOD APPLIED MECHANISM DESIGN FOR SOCIAL GOOD JOHN P DICKERSON Lecture #3 09/06/2016 CMSC828M Tuesdays & Thursdays 12:30pm 1:45pm REMINDER: SEND ME TOP 3 PRESENTATION PREFERENCES! I LL POST THE SCHEDULE TODAY

More information

6.207/14.15: Networks Lecture 24: Decisions in Groups

6.207/14.15: Networks Lecture 24: Decisions in Groups 6.207/14.15: Networks Lecture 24: Decisions in Groups Daron Acemoglu and Asu Ozdaglar MIT December 9, 2009 1 Introduction Outline Group and collective choices Arrow s Impossibility Theorem Gibbard-Satterthwaite

More information

Recap Social Choice Functions Fun Game Mechanism Design. Mechanism Design. Lecture 13. Mechanism Design Lecture 13, Slide 1

Recap Social Choice Functions Fun Game Mechanism Design. Mechanism Design. Lecture 13. Mechanism Design Lecture 13, Slide 1 Mechanism Design Lecture 13 Mechanism Design Lecture 13, Slide 1 Lecture Overview 1 Recap 2 Social Choice Functions 3 Fun Game 4 Mechanism Design Mechanism Design Lecture 13, Slide 2 Notation N is the

More information

Definitions and Proofs

Definitions and Proofs Giving Advice vs. Making Decisions: Transparency, Information, and Delegation Online Appendix A Definitions and Proofs A. The Informational Environment The set of states of nature is denoted by = [, ],

More information

3.1 Arrow s Theorem. We study now the general case in which the society has to choose among a number of alternatives

3.1 Arrow s Theorem. We study now the general case in which the society has to choose among a number of alternatives 3.- Social Choice and Welfare Economics 3.1 Arrow s Theorem We study now the general case in which the society has to choose among a number of alternatives Let R denote the set of all preference relations

More information

Mechanism Design: Basic Concepts

Mechanism Design: Basic Concepts Advanced Microeconomic Theory: Economics 521b Spring 2011 Juuso Välimäki Mechanism Design: Basic Concepts The setup is similar to that of a Bayesian game. The ingredients are: 1. Set of players, i {1,

More information

Coordination and Cheap Talk in a Battle of the Sexes with Private Information

Coordination and Cheap Talk in a Battle of the Sexes with Private Information Department of Economics Coordination and Cheap Talk in a Battle of the Sexes with Private Information Department of Economics Discussion Paper 3-0 Chirantan Ganguly Indrajit Ray Coordination and Cheap

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

Recap Social Choice Fun Game Voting Paradoxes Properties. Social Choice. Lecture 11. Social Choice Lecture 11, Slide 1

Recap Social Choice Fun Game Voting Paradoxes Properties. Social Choice. Lecture 11. Social Choice Lecture 11, Slide 1 Social Choice Lecture 11 Social Choice Lecture 11, Slide 1 Lecture Overview 1 Recap 2 Social Choice 3 Fun Game 4 Voting Paradoxes 5 Properties Social Choice Lecture 11, Slide 2 Formal Definition Definition

More information

Lecture 4. 1 Examples of Mechanism Design Problems

Lecture 4. 1 Examples of Mechanism Design Problems CSCI699: Topics in Learning and Game Theory Lecture 4 Lecturer: Shaddin Dughmi Scribes: Haifeng Xu,Reem Alfayez 1 Examples of Mechanism Design Problems Example 1: Single Item Auctions. There is a single

More information

14.770: Introduction to Political Economy Lectures 1 and 2: Collective Choice and Voting

14.770: Introduction to Political Economy Lectures 1 and 2: Collective Choice and Voting 14.770: Introduction to Political Economy Lectures 1 and 2: Collective Choice and Voting Daron Acemoglu MIT September 6 and 11, 2017. Daron Acemoglu (MIT) Political Economy Lectures 1 and 2 September 6

More information

Implementation in undominated strategies by bounded mechanisms: The Pareto Correspondence

Implementation in undominated strategies by bounded mechanisms: The Pareto Correspondence Implementation in undominated strategies by bounded mechanisms: The Pareto Correspondence Saptarshi Mukherjee Eve Ramaekers Arunava Sen September 5, 2016 Preliminary and Incomplete Abstract We show that

More information

Optimal Mechanism Design without Money

Optimal Mechanism Design without Money Optimal Mechanism Design without Money Alex Gershkov, Benny Moldovanu and Xianwen Shi January, 013 Abstract We consider the standard mechanism design environment with linear utility but without monetary

More information

Expanding Choice in School Choice

Expanding Choice in School Choice Expanding Choice in School Choice Atila Abdulkadiroğlu Yeon-Koo Che Yosuke Yasuda October 15, Northwestern Seminar 1 Introduction Traditionally, students are assigned to public schools according to where

More information

Volume 31, Issue 1. Manipulation of the Borda rule by introduction of a similar candidate. Jérôme Serais CREM UMR 6211 University of Caen

Volume 31, Issue 1. Manipulation of the Borda rule by introduction of a similar candidate. Jérôme Serais CREM UMR 6211 University of Caen Volume 31, Issue 1 Manipulation of the Borda rule by introduction of a similar candidate Jérôme Serais CREM UMR 6211 University of Caen Abstract In an election contest, a losing candidate a can manipulate

More information

Static Information Design

Static Information Design Static Information Design Dirk Bergemann and Stephen Morris Frontiers of Economic Theory & Computer Science, Becker-Friedman Institute, August 2016 Mechanism Design and Information Design Basic Mechanism

More information

Near-Potential Games: Geometry and Dynamics

Near-Potential Games: Geometry and Dynamics Near-Potential Games: Geometry and Dynamics Ozan Candogan, Asuman Ozdaglar and Pablo A. Parrilo January 29, 2012 Abstract Potential games are a special class of games for which many adaptive user dynamics

More information

Mechanism Design: Implementation. Game Theory Course: Jackson, Leyton-Brown & Shoham

Mechanism Design: Implementation. Game Theory Course: Jackson, Leyton-Brown & Shoham Game Theory Course: Jackson, Leyton-Brown & Shoham Bayesian Game Setting Extend the social choice setting to a new setting where agents can t be relied upon to disclose their preferences honestly Start

More information

Bayes Correlated Equilibrium and Comparing Information Structures

Bayes Correlated Equilibrium and Comparing Information Structures Bayes Correlated Equilibrium and Comparing Information Structures Dirk Bergemann and Stephen Morris Spring 2013: 521 B Introduction game theoretic predictions are very sensitive to "information structure"

More information

Correlated Equilibrium in Games with Incomplete Information

Correlated Equilibrium in Games with Incomplete Information Correlated Equilibrium in Games with Incomplete Information Dirk Bergemann and Stephen Morris Econometric Society Summer Meeting June 2012 Robust Predictions Agenda game theoretic predictions are very

More information

The effect of learning on membership and welfare in an International Environmental Agreement

The effect of learning on membership and welfare in an International Environmental Agreement Climatic Change (202) 0:499 505 DOI 0007/s0584-0-034-5 The effect of learning on membership and welfare in an International Environmental Agreement Larry Karp Received: 7 March 200 / Accepted: 20 April

More information

NASH IMPLEMENTATION USING SIMPLE MECHANISMS WITHOUT UNDESIRABLE MIXED-STRATEGY EQUILIBRIA

NASH IMPLEMENTATION USING SIMPLE MECHANISMS WITHOUT UNDESIRABLE MIXED-STRATEGY EQUILIBRIA NASH IMPLEMENTATION USING SIMPLE MECHANISMS WITHOUT UNDESIRABLE MIXED-STRATEGY EQUILIBRIA MARIA GOLTSMAN Abstract. This note shows that, in separable environments, any monotonic social choice function

More information

Robust mechanism design and dominant strategy voting rules

Robust mechanism design and dominant strategy voting rules Theoretical Economics 9 (2014), 339 360 1555-7561/20140339 Robust mechanism design and dominant strategy voting rules Tilman Börgers Department of Economics, University of Michigan Doug Smith Federal Trade

More information

Ph.D. Preliminary Examination MICROECONOMIC THEORY Applied Economics Graduate Program June 2016

Ph.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 information

A Characterization of Single-Peaked Preferences via Random Social Choice Functions

A Characterization of Single-Peaked Preferences via Random Social Choice Functions A Characterization of Single-Peaked Preferences via Random Social Choice Functions Shurojit Chatterji, Arunava Sen and Huaxia Zeng September 2014 Paper No. 13-2014 ANY OPINIONS EXPRESSED ARE THOSE OF THE

More information

Robust Predictions in Games with Incomplete Information

Robust Predictions in Games with Incomplete Information Robust Predictions in Games with Incomplete Information joint with Stephen Morris (Princeton University) November 2010 Payoff Environment in games with incomplete information, the agents are uncertain

More information

APPLIED MECHANISM DESIGN FOR SOCIAL GOOD

APPLIED MECHANISM DESIGN FOR SOCIAL GOOD APPLIED MECHANISM DESIGN FOR SOCIAL GOOD JOHN P DICKERSON Lecture #21 11/8/2016 CMSC828M Tuesdays & Thursdays 12:30pm 1:45pm IMPOSSIBILITY RESULTS IN VOTING THEORY / SOCIAL CHOICE Thanks to: Tuomas Sandholm

More information

Theories of justice. Harsanyi s approach, Rawls approach, Unification

Theories of justice. Harsanyi s approach, Rawls approach, Unification Theories of justice Section 6.4 in JR. The choice of one swf over another is ultimately a choice between alternative sets of ethical values. Two main approaches: Harsanyi s approach, Rawls approach, Unification

More information

Near-Potential Games: Geometry and Dynamics

Near-Potential Games: Geometry and Dynamics Near-Potential Games: Geometry and Dynamics Ozan Candogan, Asuman Ozdaglar and Pablo A. Parrilo September 6, 2011 Abstract Potential games are a special class of games for which many adaptive user dynamics

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

On the Impossibility of Black-Box Truthfulness Without Priors

On the Impossibility of Black-Box Truthfulness Without Priors On the Impossibility of Black-Box Truthfulness Without Priors Nicole Immorlica Brendan Lucier Abstract We consider the problem of converting an arbitrary approximation algorithm for a singleparameter social

More information

CPS 173 Mechanism design. Vincent Conitzer

CPS 173 Mechanism design. Vincent Conitzer CPS 173 Mechanism design Vincent Conitzer conitzer@cs.duke.edu edu Mechanism design: setting The center has a set of outcomes O that she can choose from Allocations of tasks/resources, joint plans, Each

More information

Non-Manipulable Domains for the Borda Count

Non-Manipulable Domains for the Borda Count Non-Manipulable Domains for the Borda Count Martin Barbie, Clemens Puppe * Department of Economics, University of Karlsruhe D 76128 Karlsruhe, Germany and Attila Tasnádi ** Department of Mathematics, Budapest

More information

Some Notes on Costless Signaling Games

Some Notes on Costless Signaling Games Some Notes on Costless Signaling Games John Morgan University of California at Berkeley Preliminaries Our running example is that of a decision maker (DM) consulting a knowledgeable expert for advice about

More information

Follow links for Class Use and other Permissions. For more information send to:

Follow links for Class Use and other Permissions. For more information send  to: COPYRIGHT NOTICE: Ariel Rubinstein: Lecture Notes in Microeconomic Theory is published by Princeton University Press and copyrighted, c 2006, by Princeton University Press. All rights reserved. No part

More information

The Dimensions of Consensus

The Dimensions of Consensus The Dimensions of Consensus Alex Gershkov Benny Moldovanu Xianwen Shi Hebrew and Surrey Bonn Toronto February 2017 The Dimensions of Consensus February 2017 1 / 1 Introduction Legislatures/committees vote

More information

Chapter 2. Equilibrium. 2.1 Complete Information Games

Chapter 2. Equilibrium. 2.1 Complete Information Games Chapter 2 Equilibrium Equilibrium attempts to capture what happens in a game when players behave strategically. This is a central concept to these notes as in mechanism design we are optimizing over games

More information

SYMMETRIC MECHANISM DESIGN. October 19, 2015

SYMMETRIC MECHANISM DESIGN. October 19, 2015 SYMMETRIC MECHANISM DESIGN YARON AZRIELI AND RITESH JAIN Abstract. Designers of economic mechanisms often have goals that are inconsistent with fairness. This paper studies the extent to which regulators

More information

Voting and Mechanism Design

Voting and Mechanism Design José M Vidal Department of Computer Science and Engineering, University of South Carolina March 26, 2010 Abstract Voting, Mechanism design, and distributed algorithmics mechanism design. Chapter 8. Voting

More information

WORKING PAPER. A Ranking Mechanism for Coupled Binary Decision Problems

WORKING PAPER. A Ranking Mechanism for Coupled Binary Decision Problems WORKING PAPER A Ranking Mechanism for Coupled Binary Decision Problems Kilian M. Russ 1 and Justus Winkelmann 1 1 Bonn Graduate School of Economics January 18, 2016 Abstract We propose a simple mechanism

More information

Online Appendices for Large Matching Markets: Risk, Unraveling, and Conflation

Online Appendices for Large Matching Markets: Risk, Unraveling, and Conflation Online Appendices for Large Matching Markets: Risk, Unraveling, and Conflation Aaron L. Bodoh-Creed - Cornell University A Online Appendix: Strategic Convergence In section 4 we described the matching

More information

Repeated Downsian Electoral Competition

Repeated Downsian Electoral Competition Repeated Downsian Electoral Competition John Duggan Department of Political Science and Department of Economics University of Rochester Mark Fey Department of Political Science University of Rochester

More information

Multiple Equilibria in the Citizen-Candidate Model of Representative Democracy.

Multiple Equilibria in the Citizen-Candidate Model of Representative Democracy. Multiple Equilibria in the Citizen-Candidate Model of Representative Democracy. Amrita Dhillon and Ben Lockwood This version: March 2001 Abstract De Sinopoli and Turrini (1999) present an example to show

More information

Correlated Equilibria of Classical Strategic Games with Quantum Signals

Correlated 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 information

Economics Bulletin, 2012, Vol. 32 No. 1 pp Introduction. 2. The preliminaries

Economics Bulletin, 2012, Vol. 32 No. 1 pp Introduction. 2. The preliminaries 1. Introduction In this paper we reconsider the problem of axiomatizing scoring rules. Early results on this problem are due to Smith (1973) and Young (1975). They characterized social welfare and social

More information

Social Choice. Jan-Michael van Linthoudt

Social Choice. Jan-Michael van Linthoudt Social Choice Jan-Michael van Linthoudt Summer term 2017 Version: March 15, 2018 CONTENTS Remarks 1 0 Introduction 2 1 The Case of 2 Alternatives 3 1.1 Examples for social choice rules............................

More information

The Revenue Equivalence Theorem 1

The Revenue Equivalence Theorem 1 John Nachbar Washington University May 2, 2017 The Revenue Equivalence Theorem 1 1 Introduction. The Revenue Equivalence Theorem gives conditions under which some very different auctions generate the same

More information

Mechanism Design. Christoph Schottmüller / 27

Mechanism Design. Christoph Schottmüller / 27 Mechanism Design Christoph Schottmüller 2015-02-25 1 / 27 Outline 1 Bayesian implementation and revelation principle 2 Expected externality mechanism 3 Review questions and exercises 2 / 27 Bayesian implementation

More information

Comparing School Choice Mechanisms by Interim and Ex-Ante Welfare

Comparing School Choice Mechanisms by Interim and Ex-Ante Welfare This work is distributed as a Discussion Paper by the STANFORD INSTITUTE FOR ECONOMIC POLICY RESEARCH SIEPR Discussion Paper No. 10-021 Comparing School Choice Mechanisms by Interim and Ex-Ante Welfare

More information

Hans Peters, Souvik Roy, Ton Storcken. Manipulation under k-approval scoring rules RM/08/056. JEL code: D71, D72

Hans Peters, Souvik Roy, Ton Storcken. Manipulation under k-approval scoring rules RM/08/056. JEL code: D71, D72 Hans Peters, Souvik Roy, Ton Storcken Manipulation under k-approval scoring rules RM/08/056 JEL code: D71, D72 Maastricht research school of Economics of TEchnology and ORganizations Universiteit Maastricht

More information

Efficient Repeated Implementation

Efficient Repeated Implementation Efficient Repeated Implementation Jihong Lee Yonsei University and Birkbeck College, London Hamid Sabourian University of Cambridge September 2009 Abstract This paper examines repeated implementation of

More information

Essays on Mechanism Design

Essays on Mechanism Design Essays on Mechanism Design by Douglas Scott Smith A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Economics) in The University of Michigan 2011

More information

A Game-Theoretic Analysis of Games with a Purpose

A Game-Theoretic Analysis of Games with a Purpose A Game-Theoretic Analysis of Games with a Purpose The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters. Citation Published Version

More information

Towards a General Theory of Non-Cooperative Computation

Towards a General Theory of Non-Cooperative Computation Towards a General Theory of Non-Cooperative Computation (Extended Abstract) Robert McGrew, Ryan Porter, and Yoav Shoham Stanford University {bmcgrew,rwporter,shoham}@cs.stanford.edu Abstract We generalize

More information

Ex-Ante Stable Lotteries

Ex-Ante Stable Lotteries Ex-Ante Stable Lotteries Jan Christoph Schlegel Faculty of Business and Economics, University of Lausanne, Switzerland jschlege@unil.ch Abstract We study the allocation of indivisible objects (e.g. school

More information

The effect of learning on membership and welfare in an International Environmental Agreement

The effect of learning on membership and welfare in an International Environmental Agreement The effect of learning on membership and welfare in an International Environmental Agreement Larry Karp University of California, Berkeley Department of Agricultural and Resource Economics karp@are.berkeley.edu

More information

Game Theory. Monika Köppl-Turyna. Winter 2017/2018. Institute for Analytical Economics Vienna University of Economics and Business

Game Theory. Monika Köppl-Turyna. Winter 2017/2018. Institute for Analytical Economics Vienna University of Economics and Business Monika Köppl-Turyna Institute for Analytical Economics Vienna University of Economics and Business Winter 2017/2018 Static Games of Incomplete Information Introduction So far we assumed that payoff functions

More information

Introduction. 1 University of Pennsylvania, Wharton Finance Department, Steinberg Hall-Dietrich Hall, 3620

Introduction. 1 University of Pennsylvania, Wharton Finance Department, Steinberg Hall-Dietrich Hall, 3620 May 16, 2006 Philip Bond 1 Are cheap talk and hard evidence both needed in the courtroom? Abstract: In a recent paper, Bull and Watson (2004) present a formal model of verifiability in which cheap messages

More information

Efficient Random Assignment with Cardinal and Ordinal Preferences: Online Appendix

Efficient Random Assignment with Cardinal and Ordinal Preferences: Online Appendix Efficient Random Assignment with Cardinal and Ordinal Preferences: Online Appendix James C. D. Fisher December 11, 2018 1 1 Introduction This document collects several results, which supplement those in

More information

Worst-case mechanism design with undominated strategies

Worst-case mechanism design with undominated strategies Worst-case mechanism design with undominated strategies Takuro Yamashita April 26, 2009 Abstract We consider a way to evaluate mechanisms without assuming mutual knowledge of rationality among the agents.

More information

Partial lecture notes THE PROBABILITY OF VIOLATING ARROW S CONDITIONS

Partial lecture notes THE PROBABILITY OF VIOLATING ARROW S CONDITIONS Partial lecture notes THE PROBABILITY OF VIOLATING ARROW S CONDITIONS 1 B. Preference Aggregation Rules 3. Anti-Plurality a. Assign zero points to a voter's last preference and one point to all other preferences.

More information

DEPARTMENT OF ECONOMICS IMPLEMENTATION IN MIXED NASH EQUILIBRIUM

DEPARTMENT OF ECONOMICS IMPLEMENTATION IN MIXED NASH EQUILIBRIUM DEPARTMENT OF ECONOMICS IMPLEMENTATION IN MIXED NASH EQUILIBRIUM Claudio Mezzetti, University of Warwick, UK Ludovic Renou, University of Leicester, UK Working Paper No. 09/10 April 2009 Updated January

More information

UC Berkeley Haas School of Business Game Theory (EMBA 296 & EWMBA 211) Summer Social learning and bargaining (axiomatic approach)

UC Berkeley Haas School of Business Game Theory (EMBA 296 & EWMBA 211) Summer Social learning and bargaining (axiomatic approach) UC Berkeley Haas School of Business Game Theory (EMBA 296 & EWMBA 211) Summer 2015 Social learning and bargaining (axiomatic approach) Block 4 Jul 31 and Aug 1, 2015 Auction results Herd behavior and

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

Graph Theoretic Characterization of Revenue Equivalence

Graph Theoretic Characterization of Revenue Equivalence Graph Theoretic Characterization of University of Twente joint work with Birgit Heydenreich Rudolf Müller Rakesh Vohra Optimization and Capitalism Kantorovich [... ] problems of which I shall speak, relating

More information

Algorithmic Game Theory and Applications

Algorithmic Game Theory and Applications Algorithmic Game Theory and Applications Lecture 18: Auctions and Mechanism Design II: a little social choice theory, the VCG Mechanism, and Market Equilibria Kousha Etessami Reminder: Food for Thought:

More information

Turnout, Polarization, and Duverger s Law.

Turnout, Polarization, and Duverger s Law. Turnout, Polarization, and Duverger s Law. Online Appendix: Steven Callander & Catherine H. Wilson November 0, 006 Proof of Lemma 1: The assumptions on f imply for e>d that V (E) 0 as e e the interval

More information

Selecting Efficient Correlated Equilibria Through Distributed Learning. Jason R. Marden

Selecting Efficient Correlated Equilibria Through Distributed Learning. Jason R. Marden 1 Selecting Efficient Correlated Equilibria Through Distributed Learning Jason R. Marden Abstract A learning rule is completely uncoupled if each player s behavior is conditioned only on his own realized

More information

Crowdsourcing contests

Crowdsourcing contests December 8, 2012 Table of contents 1 Introduction 2 Related Work 3 Model: Basics 4 Model: Participants 5 Homogeneous Effort 6 Extensions Table of Contents 1 Introduction 2 Related Work 3 Model: Basics

More information

Robust Mechanism Design and Robust Implementation

Robust Mechanism Design and Robust Implementation Robust Mechanism Design and Robust Implementation joint work with Stephen Morris August 2009 Barcelona Introduction mechanism design and implementation literatures are theoretical successes mechanisms

More information

Second Welfare Theorem

Second Welfare Theorem Second Welfare Theorem Econ 2100 Fall 2015 Lecture 18, November 2 Outline 1 Second Welfare Theorem From Last Class We want to state a prove a theorem that says that any Pareto optimal allocation is (part

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

Efficient Random Assignment under a Combination of Ordinal and Cardinal Information on Preferences

Efficient Random Assignment under a Combination of Ordinal and Cardinal Information on Preferences Fondazione Eni Enrico Mattei Working Papers 2-21-2011 Efficient Random Assignment under a Combination of Ordinal and Cardinal Information on Preferences Stergios Athanassoglou athanassoglou@gmail.com Follow

More information

Redistribution Mechanisms for Assignment of Heterogeneous Objects

Redistribution Mechanisms for Assignment of Heterogeneous Objects Redistribution Mechanisms for Assignment of Heterogeneous Objects Sujit Gujar Dept of Computer Science and Automation Indian Institute of Science Bangalore, India sujit@csa.iisc.ernet.in Y Narahari Dept

More information

Basic Game Theory. Kate Larson. January 7, University of Waterloo. Kate Larson. What is Game Theory? Normal Form Games. Computing Equilibria

Basic Game Theory. Kate Larson. January 7, University of Waterloo. Kate Larson. What is Game Theory? Normal Form Games. Computing Equilibria Basic Game Theory University of Waterloo January 7, 2013 Outline 1 2 3 What is game theory? The study of games! Bluffing in poker What move to make in chess How to play Rock-Scissors-Paper Also study of

More information

Theory of Mechanism Design

Theory of Mechanism Design Theory of Mechanism Design Debasis Mishra 1 October 28, 2015 1 Economics and Planning Unit, Indian Statistical Institute, 7 Shahid Jit Singh Marg, New Delhi 110016, India, E-mail: dmishra@isid.ac.in 2

More information

EC3224 Autumn Lecture #04 Mixed-Strategy Equilibrium

EC3224 Autumn Lecture #04 Mixed-Strategy Equilibrium Reading EC3224 Autumn Lecture #04 Mixed-Strategy Equilibrium Osborne Chapter 4.1 to 4.10 By the end of this week you should be able to: find a mixed strategy Nash Equilibrium of a game explain why mixed

More information

Utilitarianism, Degressive Proportionality and the Constitution of a Federal Assembly

Utilitarianism, Degressive Proportionality and the Constitution of a Federal Assembly 1 Utilitarianism, Degressive Proportionality and the Constitution of a Federal Assembly Luc Bovens and Stephan Hartmann 1 Introduction. A federal assembly consists of a number of representatives for each

More information

Social Choice Theory. Felix Munoz-Garcia School of Economic Sciences Washington State University. EconS Advanced Microeconomics II

Social Choice Theory. Felix Munoz-Garcia School of Economic Sciences Washington State University. EconS Advanced Microeconomics II Social Choice Theory Felix Munoz-Garcia School of Economic Sciences Washington State University EconS 503 - Advanced Microeconomics II Social choice theory MWG, Chapter 21. JR, Chapter 6.2-6.5. Additional

More information

Economic Core, Fair Allocations, and Social Choice Theory

Economic Core, Fair Allocations, and Social Choice Theory Chapter 9 Nathan Smooha Economic Core, Fair Allocations, and Social Choice Theory 9.1 Introduction In this chapter, we briefly discuss some topics in the framework of general equilibrium theory, namely

More information

Some Notes on Adverse Selection

Some Notes on Adverse Selection Some Notes on Adverse Selection John Morgan Haas School of Business and Department of Economics University of California, Berkeley Overview This set of lecture notes covers a general model of adverse selection

More information

Appendix of Homophily in Peer Groups The Costly Information Case

Appendix of Homophily in Peer Groups The Costly Information Case Appendix of Homophily in Peer Groups The Costly Information Case Mariagiovanna Baccara Leeat Yariv August 19, 2012 1 Introduction In this Appendix we study the information sharing application analyzed

More information

Non-Existence of Equilibrium in Vickrey, Second-Price, and English Auctions

Non-Existence of Equilibrium in Vickrey, Second-Price, and English Auctions Non-Existence of Equilibrium in Vickrey, Second-Price, and English Auctions Matthew O. Jackson September 21, 2005 Forthcoming: Review of Economic Design Abstract A simple example shows that equilibria

More information

Microeconomics II Lecture 4: Incomplete Information Karl Wärneryd Stockholm School of Economics November 2016

Microeconomics II Lecture 4: Incomplete Information Karl Wärneryd Stockholm School of Economics November 2016 Microeconomics II Lecture 4: Incomplete Information Karl Wärneryd Stockholm School of Economics November 2016 1 Modelling incomplete information So far, we have studied games in which information was complete,

More information

Chapter 2. Equilibrium. 2.1 Complete Information Games

Chapter 2. Equilibrium. 2.1 Complete Information Games Chapter 2 Equilibrium The theory of equilibrium attempts to predict what happens in a game when players behave strategically. This is a central concept to this text as, in mechanism design, we are optimizing

More information

Contracts under Asymmetric Information

Contracts under Asymmetric Information Contracts under Asymmetric Information 1 I Aristotle, economy (oiko and nemo) and the idea of exchange values, subsequently adapted by Ricardo and Marx. Classical economists. An economy consists of a set

More information

Static Information Design

Static Information Design Static nformation Design Dirk Bergemann and Stephen Morris European Summer Symposium in Economic Theory, Gerzensee, July 2016 Mechanism Design and nformation Design Mechanism Design: Fix an economic environment

More information

Revenue Maximization in Multi-Object Auctions

Revenue Maximization in Multi-Object Auctions Revenue Maximization in Multi-Object Auctions Benny Moldovanu April 25, 2017 Revenue Maximization in Multi-Object Auctions April 25, 2017 1 / 1 Literature Myerson R. (1981): Optimal Auction Design, Mathematics

More information

SEQUENTIAL EQUILIBRIA IN BAYESIAN GAMES WITH COMMUNICATION. Dino Gerardi and Roger B. Myerson. December 2005

SEQUENTIAL EQUILIBRIA IN BAYESIAN GAMES WITH COMMUNICATION. Dino Gerardi and Roger B. Myerson. December 2005 SEQUENTIAL EQUILIBRIA IN BAYESIAN GAMES WITH COMMUNICATION By Dino Gerardi and Roger B. Myerson December 2005 COWLES FOUNDATION DISCUSSION AER NO. 1542 COWLES FOUNDATION FOR RESEARCH IN ECONOMICS YALE

More information

Incentive-Compatible Voting Rules with Positively Correlated Beliefs

Incentive-Compatible Voting Rules with Positively Correlated Beliefs Incentive-Compatible Voting Rules with Positively Correlated Beliefs Mohit Bhargava, Dipjyoti Majumdar and Arunava Sen August 13, 2014 Abstract We study the consequences of positive correlation of beliefs

More information

Epsilon Ex Post Implementation

Epsilon Ex Post Implementation Epsilon Ex Post Implementation Mehmet Barlo Nuh Aygun Dalkiran February 10, 2014 Abstract We provide necessary and sufficient conditions for epsilon ex post implementation. Our analysis extends Bergemann

More information

Strategies under Strategic Uncertainty

Strategies under Strategic Uncertainty Discussion Paper No. 18-055 Strategies under Strategic Uncertainty Helene Mass Discussion Paper No. 18-055 Strategies under Strategic Uncertainty Helene Mass Download this ZEW Discussion Paper from our

More information

Efficiency and Stability of Probabilistic Assignments in Marriage Problems

Efficiency and Stability of Probabilistic Assignments in Marriage Problems Efficiency and Stability of Probabilistic Assignments in Marriage Problems Battal Doğan Kemal Yıldız March 23, 205 Abstract We study marriage problems where two groups of agents, men and women, match each

More information

Microeconomic Theory (501b) Problem Set 10. Auctions and Moral Hazard Suggested Solution: Tibor Heumann

Microeconomic Theory (501b) Problem Set 10. Auctions and Moral Hazard Suggested Solution: Tibor Heumann Dirk Bergemann Department of Economics Yale University Microeconomic Theory (50b) Problem Set 0. Auctions and Moral Hazard Suggested Solution: Tibor Heumann 4/5/4 This problem set is due on Tuesday, 4//4..

More information

Information Design. Dirk Bergemann and Stephen Morris. Johns Hopkins University April 2017

Information Design. Dirk Bergemann and Stephen Morris. Johns Hopkins University April 2017 Information Design Dirk Bergemann and Stephen Morris Johns Hopkins University April 2017 Mechanism Design and Information Design Basic Mechanism Design: Fix an economic environment and information structure

More information

BAYES CORRELATED EQUILIBRIUM AND THE COMPARISON OF INFORMATION STRUCTURES IN GAMES. Dirk Bergemann and Stephen Morris

BAYES CORRELATED EQUILIBRIUM AND THE COMPARISON OF INFORMATION STRUCTURES IN GAMES. Dirk Bergemann and Stephen Morris BAYES CORRELATED EQUILIBRIUM AND THE COMPARISON OF INFORMATION STRUCTURES IN GAMES By Dirk Bergemann and Stephen Morris September 203 Revised October 204 COWLES FOUNDATION DISCUSSION PAPER NO. 909RR COWLES

More information