The value of a stochastic information structure

Upon observing a signal, a Bayesian decision maker updates her probability distribution over the state space, chooses an action, and receives a payoff that depends on the state and the action taken. An information structure determines the set of possible signals and the probability of each signal given a state. For a fixed decision problem, the value of an information structure is the maximal expected utility that the decision maker can get when the observed signals are governed by this structure. Thus, every decision problem induces a preference order over information structures according to their value. We characterize preference orders that can be obtained in this way. We also characterize the functions defined over information structures that measure their value.     

Learning to play games in extensive form by valuation

Game theoretic models of learning which are based on the strategic form of the game cannot explain learning in games with large extensive form. We study learning in such games by using valuation of moves. A valuation for a player is a numeric assessment of her moves that purports to reflect their desirability. We consider a myopic player, who chooses moves with the highest valuation. Each time the game is played, the player revises her valuation by assigning the payoff obtained in the play to each of the moves she has made. We show for a repeated win-lose game that if the player has a winning strategy in the stage game, there is almost surely a time after which she always wins. When a player has more than two payoffs, a more elaborate learning procedure is required. We consider one that associates with each move the average payoff in the rounds in which this move was made. When all players adopt this learning procedure, with some perturbations, then, with probability 1 there is a time after which strategies that are close to subgame perfect equilibrium are played. A single player who adopts this procedure can guarantee only her individually rational payoff.     

Induced Preferences, Nonadditive Beliefs, and Multiple Priors

We study a decision maker who follows the Savage axioms. We show that if he or she is able to take unobservable actions that influence the probabilities of outcomes, then it can appear to an outsider as if his or her subjective probabilities are nonadditive. Implications for multiperiod decision are explored. We extend the model to include a second individual who is also able to take a hidden action. We show that this may induce uncertainty-averse preferences over some class of acts, even if the second individual acts to help the decision maker with high probability.     

Choosing the two finalists

This paper studies a decision maker who for each choice set selects a subset of (at most) two alternatives. We axiomatize three types of procedures: (i) The top two: the decision maker has in mind an ordering and chooses the two maximal alternatives. (ii) The two extremes: the decision maker has in mind an ordering and chooses the maximal and the minimal alternatives. (iii) The top and the top: the decision maker has in mind two orderings and he chooses the maximal element from each.     

Proper scoring rules for general decision models

If a decision maker whose behavior conforms to the max-min expected utility model is faced with a scoring rule for a subjective expected utility decision maker, she will always announce a probability belonging to her set of priors; moreover, for any prior in the set, there is a scoring rule inducing the agent to announce that prior. We also show that on the domain of Choquet expected utility preferences with risk neutral lottery evaluation and totally monotone capacities, proper scoring rules do not exist. This implies the non-existence of proper scoring rules for any larger class of preferences (CEU with convex capacities, multiple priors).     

Aggregation of multiple prior opinions

Experts are asked to provide their advice in a situation of uncertainty. They adopt the decision maker?s utility function, but each has a potentially different set of prior probabilities, and so does the decision maker. The decision maker and the experts maximize the minimal expected utility with respect to their sets of priors. We show that a natural Pareto condition is equivalent to the existence of a set Λ of probability vectors over the experts, interpreted as possible allocations of weights to the experts, such that (i) the decision maker?s set of priors is precisely all the weighted-averages of priors, where an expert?s prior is taken from her set and the weight vector is taken from Λ; (ii) the decision maker?s valuation of an act is the minimal weighted valuation, over all weight vectors in Λ, of the experts? valuations.     

Revealed preferences and aspirations in warm glow theory

In warm glow models, an agent may prefer one alternative but aspire to choose another. The agent chooses her aspiration if she gets a sufficiently large warm glow payoff for acting as she aspires. This basic framework is widely used in models of turnout in elections and contributions to public goods, but is often criticized for being ad hoc. In this paper, we provide choice-theoretic foundations for warm glow theory. We characterize the empirical content of warm glow theory, show how to infer the core elements of the model from data and show that it is possible to predict behavior even when preferences and aspirations are not revealed. Our results provide support for assumptions often made in the literature and suggest new applications for warm glow models.     

Knightian decision theory and econometric inferences

An uncertainty averse Knightian decision maker has a set of probability distributions over outcomes and chooses something other than the status quo only if the change increases the expected payoff according to all the distributions. It is possible to define a standardized degree of uncertainty aversion. To each such degree, there corresponds a set of prior distributions over the parameters of a Gaussian linear regression model, these priors being centered on a uniform prior. The set of posterior means corresponding to this set of priors has the same properties as a standard confidence region.     

Minimax regret with imperfect ex-post knowledge of the state

We consider decision problems under complete ignorance and extend the minimax regret principle to situations where, after taking an action, the decision maker does not necessarily learn the state of the world. For example, if the decision maker only learns what the outcome is, then all she knows is that the actual state is one of the (possibly several) states that yield the observed outcome under the chosen action. We refer to this situation as imperfect ex-post information. We show that, given a choice between more information and less information, the decision maker prefers the latter. We also extend the framework to encompass the possibility of less than the extreme degree of pessimism that characterizes the minimax regret criterion.     

Directed Networks with Spillovers

We study noncooperative network formation in two types of directed networks. In the first type, called the model with global spillovers, the payoff of a player depends on the number of links she forms as well as the total number of links formed by all other players. In the second type, called the model with local spillovers, the payoff of a player depends on the number of links she forms and the total number of links formed by her immediate neighbors, as well as the number of links formed by players outside her neighborhood. For both classes of games we investigate the existence of pure strategy Nash equilibria and characterize the Nash networks under a number of different second order conditions on the payoff function.     

Redistributive Taxation under Ethical Behaviour*

We consider the implications of ethical behaviour on the effect of a redistributive tax‐transfer system. In choosing their labour supplies, individuals take into account whether their tax liabilities correspond to what they view as ethically acceptable. If tax liabilities are viewed as ethically acceptable, a taxpayer behaves ethically, does not distort her behaviour, and chooses to work as if she were not taxed. On the other hand, if ethical behaviour results in tax liabilities that exceed those that are ethically acceptable, she behaves egoistically (partially or fully), distorts her behaviour, and chooses her labour supply taking into account the income tax. We establish taxpayers' equilibrium behaviour and obtain that labour supply is less elastic when taxpayers may behave ethically than when they act egoistically. We characterise and compare the egoistic voting equilibrium linear tax schedules under potentially ethical and egoistic behaviour. We also compare our results to those obtained under altruism, an alternative benchmark.     

Utility from anticipation and personal equilibrium

I develop a dynamic model of individual decisionmaking in which the agent derives utility from physical outcomes as well as from rational beliefs about physical outcomes (“anticipation”), and these two payoff components can interact. Beliefs and behavior are jointly determined in a personal equilibrium by the requirement that behavior given past beliefs must be consistent with those beliefs. I explore three phenomena made possible by utility from anticipation, and prove that if the decisionmaker’s behavior is distinguishable from a person’s who cares only about physical outcomes, she must exhibit at least one of these phenomena. First, the decisionmaker can be prone to self-fulfilling expectations. Second, she might be time-inconsistent even if her preferences in all periods are identical. Third, she might exhibit informational preferences, where these preferences are intimately connected to her attitudes toward disappointments. Applications of the framework to reference-dependent preferences, impulsive behaviors, and emotionally difficult choices are discussed.     

SELLING A LEMON UNDER DEMAND UNCERTAINTY

We consider a seller who has private information about the quality of her good but is uncertain about buyer arrivals. Assuming that the high‐quality seller insists on a price, we show that the low‐quality seller's surplus and pricing strategy crucially depend on buyers' knowledge about the demand state. If they are also uncertain about demand, then demand uncertainty increases the low‐quality seller's expected payoff, and her optimal strategy is to lower the price after some time. If buyers know the demand state, then demand uncertainty does not affect the low‐quality seller's payoff, but she must employ a sophisticated pricing strategy.     

Choquet Rationality

Consider a decision problem under uncertainty for a decision maker with known (utility) payoffs over prizes. We say that an act is Choquet (Shafer, Bernoulli) rational if for some capacity (belief function, probability) over the set of states, it maximizes her “expected” utility. We show that an act may be Choquet rational without being Bernoulli rational, but it is Choquet rational if and only if it is Shafer rational. Journal of Economic Literature Classification Numbers: C72, D81.     

The subjective expected utility hypothesis and revealed preference

Summary A choice behavior of a decision maker is said to satisfy the subjective expected utility hypothesis if there exist a utility and a subjective probability such that a decision maker chooses acts with the highest expected utility. We present a revealed preference characterization of choice behavior that is consistent with the subjective expected utility hypothesis. Our result applies to any state space and compact sets of prizes and observations (either finite or infinite).     

The rationality and efficacy of decisions under uncertainty and the value of an experiment

Summary A decision maker faces a known prior distribution over payoff relevant states. We compare the expected utility of this individual under two scenarios. In the first, the decision maker makes a choice without further information. In the second, the decision maker has access to an experiment before choosing an action. However, the decision maker does not know the true joint distribution over states and messages. The value of the experiment as measured by the difference in the two utility levels can be negative as well as positive. We give a condition which is necessary and sufficient for the experiment to be valuable in our sense, for any decision problem.An earlier version of this paper was circulated under the title Noisy Bayes Updating and the Value of Information. We have gained from the comments of Stephen Coate, John Geanakoplos, Larry Samuelson, Timothy Van Zandt and seminar participants at Harvard Business School, Princeton, Boston University, the international conference on game theory at Stony Brook 1992 and the Winter meeting of the Econometric Society at Anaheim 1993. The first author received support for this project from NSF grant #SES-9308515 and a University of Pennsylvania Research Foundation Grant.     

Optimal agenda-setter timing

Abstract .  We analyze the optimal timing problem of an agenda setter who can propose a project that is then voted on by a committee. The payoff consists of a common stochastic component and an idiosyncratic component that differs among committee members. The agenda setter may be biased for or against the project, relative to the median committee member, and chooses when to call the vote. We analyze how the timing decision depends on the political environment and the bias of the agenda setter. We show that both positively and negatively biased agenda setters can implement decisions that differ from those preferred by the median committee member.     

Cycling in a stochastic learning algorithm for normal form games

In this paper we study a stochastic learning model for 2×2 normal form games that are played repeatedly. The main emphasis is put on the emergence of cycles. We assume that the players have neither information about the payoff matrix of their opponent nor about their own. At every round each player can only observe his or her action and the payoff he or she receives. We prove that the learning algorithm, which is modeled by an urn scheme proposed by Arthur (1993), leads with positive probability to a cycling of strategy profiles if the game has a mixed Nash equilibrium. In case there are strict Nash equilibria, the learning process converges a.s. to the set of Nash equilibria.     

How complex are networks playing repeated games?

Summary. This paper examines implications of complexity cost in implementing repeated game strategies through networks with finitely many classifiers. A network consists of individual classifiers that summarize the history of repeated play according to a weighted sum of the empirical frequency of the outcomes of the stage game, and a decision unit that chooses an action in each period based on the summaries of the classifiers. Each player maximizes his long run average payoff, while minimizing the complexity cost of implementing his strategy through a network, measured by its number of classifiers. We examine locally stable equilibria where the selected networks are robust against small perturbations. In any locally stable equilibrium, no player uses a network with more than a single classifier. Moreover, the set of locally stable equilibrium payoff vectors lies on two line segments in the payoff space of the stage game. Received: May 9, 1997; revised version: November 18, 1997     

A Bayesian Approach to Uncertainty Aversion

The Ellsberg paradox demonstrates that people's beliefs over uncertain events might not be representable by subjective probability. We show that if a risk averse decision maker, who has a well defined Bayesian prior, perceives an Ellsberg type decision problem as possibly composed of a bundle of several positively correlated problems, she will be uncertainty averse. We generalize this argument and derive sufficient conditions for uncertainty aversion.