Learning by trial and error

A person learns by trial and error if he occasionally tries out new strategies, rejecting choices that are erroneous in the sense that they do not lead to higher payoffs. In a game, however, strategies can become erroneous due to a change of behavior by someone else. We introduce a learning rule in which behavior is conditional on whether a player experiences an error of the first or second type. This rule, called interactive trial and error learning, implements Nash equilibrium behavior in any game with generic payoffs and at least one pure Nash equilibrium.     

Substantive Rationality and Backward Induction

S. Modica and A. Rustichini (1994, Theory and Decision37, 107–124) provided a logic for reasoning about knowledge where agents may be unaware of certain propositions. However, their original approach had the unpleasant property that nontrivial unawareness was incompatible with partitional information structures. More recently, Modica and Rustichini (1999, Games Econ. Behav.27, 265–298) have provided an approach that allows for nontrivial unawareness in partitional information structures. Here it is shown that their approach can be viewed as a special case of a general approach to unawareness considered by R. Fagin and J. Y. Halpern (1988, Artificial Intelligence34, 39–76). Journal of Economic Literature Classification Numbers: D80, D83.     

Maximal Domain of Preferences in the Division Problem

The division problem consists of allocating an amount of a perfectly divisible good among a group of n agents. Sprumont (1991) showed that if agents have single-peaked preferences over their shares, then the uniform allocation rule is the unique strategy-proof, efficient, and anonymous rule. We identify the maximal set of preferences, containing the set of single-peaked preferences, under which there exists at least one rule satisfying the properties of strategy-proofness, efficiency, and strong symmetry. In addition, we show that our characterization implies a slightly weaker version of Ching and Serizawa's (1998) result. Journal of Economic Literature Classification Numbers: D71, D78, D63.     

STRATEGY‐PROOF ALLOCATION OF INDIVISIBLE GOODS AMONG COUPLES

We consider the problem of allocating indivisible goods among couples. Agents in a couple share the indivisible good assigned to them. The main result is that an allocation rule is strategy‐proof, neutral and non‐bossy if and only if it is serially dictatorial. An allocation rule is serially dictatorial if there is a priority order of couples and a function that identifies who chooses in each couple, such that for all preference profiles, a good assigned to couple i is the best element according to the preference of the identified agent in couple i among the remaining goods when the couples with higher priorities have made their choice.     

Fully sincere voting

In a general social choice framework where the requirement of strategy-proofness may not be sensible, we call a social choice rule fully sincere if it never gives any individual an incentive to vote for a less-preferred alternative over a more-preferred one and provides an incentive to vote for an alternative if and only if it is preferred to the default option that would result from abstaining. If the social choice rule can depend only on the number of votes that each alternative receives, those rules satisfying full sincerity are convex combinations of the rule that chooses each alternative with probability equal to the proportion of the vote it receives and an arbitrary rule that ignores voters' preferences. We note a sense in which the natural probabilistic analog of approval voting is the fully sincere rule that allows voters maximal flexibility in expressing their preferences and gives these preferences maximal weight.     

Rank-order tournaments and selection

Rank-order tournaments are often presented as devices for aligning incentives in a principal-agent setting. In most of this literature agents are expected to be identical so that the principal is indifferent ex ante as to who wins the contest, implying that the selection properties of the tournament can be ignored. In this paper we consider a tournament which is not necessarily symmetric, and in which agent type is private information. The principal cares about who wins, but the basic tournament will not achieve perfect selection; the lower-type agent may sometimes win. In a two-player tournament we present a simple reward system in which the winner's reward depends upon which (if any) of two “test standards” is passed; conditions are presented under which this system allows the principal to choose the best agent. This system can be extended in a simple manner to rank types in ann-player tournament. We suggest that the theory can be applied to internal labor markets and research contests.     

Stationary equilibrium transition rules for an overlapping generations model with uncertainty

An overlapping generations model incorporating random production shocks is studied. Households have finite life spans. Futures markets are incomplete. Agents have full information in one case, and receive only a limited signal in another. In both instances the existence of a time-autonomous transition rule is proved such that if all agents forecast using it, the economy's actual growth will bear the predictions out. The rule corresponds to a steady state for a non-stochastic model. With limited information, it is seen to depend on all past signals. Several approximation theorems which may facilitate future applications are presented.     

Committee Design with Endogenous Information

Identical agents gather costly information, and then aggregate it through voting. Because information is a public good, information is underprovided relative to the social optimum. A "good" voting rule must give incentives to acquire information, as well as aggregate information efficiently. A voting rule that requires a large plurality (in the extreme, unanimity) to upset the status quo can be optimal only if the information available to each agent is sufficiently accurate. This result is independent of the preferences of voters and of the cost of information.     

Incentives in the probabilistic serial mechanism

The probabilistic serial mechanism (Bogomolnaia and Moulin, 2001 [9]) is ordinally efficient but not strategy-proof. We study incentives in the probabilistic serial mechanism for large assignment problems. We establish that for a fixed set of object types and an agent with a given expected utility function, if there are sufficiently many copies of each object type, then reporting ordinal preferences truthfully is a weakly dominant strategy for the agent (regardless of the number of other agents and their preferences). The non-manipulability and the ordinal efficiency of the probabilistic serial mechanism support its implementation instead of random serial dictatorship in large assignment problems.     

Weak monotonicity and Bayes–Nash incentive compatibility

An allocation rule is called Bayes–Nash incentive compatible, if there exists a payment rule, such that truthful reports of agents' types form a Bayes–Nash equilibrium in the direct revelation mechanism consisting of the allocation rule and the payment rule. This paper provides a characterization of Bayes–Nash incentive compatible allocation rules in social choice settings where agents have multi-dimensional types, quasi-linear utility functions and interdependent valuations. The characterization is derived by constructing complete directed graphs on agents' type spaces with cost of manipulation as lengths of edges. Weak monotonicity of the allocation rule corresponds to the condition that all 2-cycles in these graphs have non-negative length. For the case that type spaces are convex and the valuation for each outcome is a linear function in the agent's type, we show that weak monotonicity of the allocation rule together with an integrability condition is a necessary and sufficient condition for Bayes–Nash incentive compatibility.     

Weak monotonicity and Bayes–Nash incentive compatibility

An allocation rule is called Bayes–Nash incentive compatible, if there exists a payment rule, such that truthful reports of agents' types form a Bayes–Nash equilibrium in the direct revelation mechanism consisting of the allocation rule and the payment rule. This paper provides a characterization of Bayes–Nash incentive compatible allocation rules in social choice settings where agents have multi-dimensional types, quasi-linear utility functions and interdependent valuations. The characterization is derived by constructing complete directed graphs on agents' type spaces with cost of manipulation as lengths of edges. Weak monotonicity of the allocation rule corresponds to the condition that all 2-cycles in these graphs have non-negative length. For the case that type spaces are convex and the valuation for each outcome is a linear function in the agent's type, we show that weak monotonicity of the allocation rule together with an integrability condition is a necessary and sufficient condition for Bayes–Nash incentive compatibility.     

Group identification with (incomplete) preferences

We consider the problem of identifying members of a group based on individual opinions. Since agents do not have preferences in the model, properties of rules that concern preferences (e.g., strategy‐proofness and efficiency) have not been studied in the literature. We fill this gap by working with a class of incomplete preferences derived directly from opinions. Our main result characterizes a new family of group identification rules, called voting‐by‐equitable‐committees rules, using two well‐known properties: strategy‐proofness and equal treatment of equals. Our family contains as a special case the consent rules (Samet & Schmeidler. J. Econ. Theory, 110 (2003), pp. 213–233), which are symmetric and embody various degrees of liberalism and democracy; and it also includes dictatorial and oligarchic rules that value agents’ opinions differently. In the presence of strategy‐proofness, efficiency turns out to be equivalent to non‐degeneracy (i.e., any agent may potentially be included or excluded from the group). This implies that a rule satisfies strategy‐proofness, efficiency, and equal treatment of equals if, and only if, it is a non‐degenerate voting‐by‐equitable‐committees rule.     

PRICE‐LEVEL DETERMINATION UNDER DISPERSED INFORMATION AND MONETARY POLICY*

This paper considers the determination of aggregate price level under dispersed information. A Central Bank sets policy in response to its noisy measure of the price level, and each agent makes its decisions by observing a subset of data. Information revealed to the agents and the bank is determined endogenously. It is shown that the aggregate state of the economy is not revealed perfectly to anybody but this economy behaves as if it is a representative‐agent economy in which the representative agent has perfect information while the Bank has partial information. The Bank's information set affects fluctuations in the price level through its effect on policy.     

Reciprocal Supervision,Collusion and Organizational Design

We study the design of supervisory functions in an organization with one principal and two agents. Each agent can perform supervision activities regarding the other agent. We characterize the way the principal must structure incentive payments to avoid any collusive activity between agents. In particular, it is shown that better mutual information between agents may hurt the principal. The other main result is the possibility that it may be better to give up one supervisory function or to have a third party be the supervisor if possible. Finally, we show that such a dual supervisory structure raises the possibility that letting collusion happen may be the best policy.     

Adaptive play with spatial sampling

We consider a 2×2 coordination game where each agent interacts with his neighbors on a ring. Ellison (1993, Econometrica 61, 1047–1071) shows that the discrete dynamical system generated by the myopic best-reply rule converges to a Nash equilibrium or to a two-period limit cycle. Following Young (1993, Econometrica 61, 57–84), we consider a best-reply process with a sampling procedure. Particularly, we introduce a spatial sampling procedure: each agent observes a sample of information in his neighborhood and plays a best reply to it. We show that if the size of the sample of information is not too large, the best-reply process converges almost surely to a Nash equilibrium. If in addition agents experiment with small probabilities, we show that, in most cases, the risk-dominant equilibrium prevails in the long run. Furthermore, it turns out that the convergence is faster than in Ellison.     

Individual accountability in teams

We consider a model of team production in which the principal observes only the team output, but agents can monitor one another (at a cost) and provide reports to the principal. We consider the problem faced by a principal who is prevented from penalizing an agent without evidence showing that the agent failed to complete his assigned actions. We show the first-best (high effort but no monitoring) can be achieved, but only if the principal assigns second-best actions. The principal requires monitoring, but agents do not monitor, and as long as output is high, the principal does not penalize agents who fail to monitor. If the principal has the responsibility for monitoring, the first-best outcome cannot be achieved, thus we identify an incentive for delegated monitoring even when agents have no informational advantage.     

Fully sincere voting

In a general social choice framework where the requirement of strategy-proofness may not be sensible, we call a social choice rule fully sincere if it never gives any individual an incentive to vote for a less-preferred alternative over a more-preferred one and provides an incentive to vote for an alternative if and only if it is preferred to the default option that would result from abstaining. If the social choice rule can depend only on the number of votes that each alternative receives, those rules satisfying full sincerity are convex combinations of the rule that chooses each alternative with probability equal to the proportion of the vote it receives and an arbitrary rule that ignores voters' preferences. We note a sense in which the natural probabilistic analog of approval voting is the fully sincere rule that allows voters maximal flexibility in expressing their preferences and gives these preferences maximal weight.     

Bankruptcy as a Control Device in Economies in Transition

We examine in this paper the design of a liquidation, or bankruptcy, policy in a partially centralized economy characterized by imperfect information. We employ a two-period model to analyze the effects of an optimal liquidation rule on the efficiency of resource allocation and choice of managerial effort when managers have private information about effort and firm productivity. First-period investment is used by the regulator to discipline the manager and to extract information. The tradeoff between disciplinary effect and information extraction might be best solved by implementing inefficient liquidation policies. Inefficiencies in liquidation policies can occur even if the regulator believes that he is facing a given type of firm with probability close to one. J. Comp. Econom., June 1995, 20(3), pp. 265-301. Cornell University, Ithaca, New York 14853-7601.     

An efficiency characterization of plurality rule in collective choice problems

We consider the problem of selecting alternatives from a set of feasible alternatives over which each agent is endowed with a strict preference. We show that there is one and only one rule that satisfies anonymity, neutrality, efficiency, tops-only, and reinforcement. The rule is known as plurality rule, which selects the alternative(s) most preferred by the largest number of agents. I would like to thank William Thomson for helpful suggestions and discussions. I am grateful to Biung-Ghi Ju, Hyungjun Kim, and Yan-An Hwang for detailed comments. I am also indebted to the Editor and an anonymous referee for valuable suggestions. As usual, I am responsible for any remaining deficiency.     

Probabilistic assignment of objects: Characterizing the serial rule

We study the problem of assigning a set of objects to a set of agents, when each agent receives one object and has strict preferences over the objects. In the absence of monetary transfers, we focus on the probabilistic rules, which take the ordinal preferences as input. We characterize the serial rule, proposed by Bogomolnaia and Moulin (2001) [2]: it is the only rule satisfying sd efficiency, sd no-envy, and bounded invariance. A special representation of feasible assignment matrices by means of consumption processes is the key to the simple and intuitive proof of our main result.