Paul J. Healy

The Trouble With Overconfidence

This paper presents a reconciliation of the three distinct ways in which the research literature has defined overconfidence: (1) overestimation of ones actual performance, (2) overplacement of ones performance relative to others, and (3) excessive precision in ones beliefs. Experimental evidence shows that reversals of the first two (apparent underconfidence), when they occur, tend to be on different types of tasks. On difficult tasks, people overestimate their actual performances but also believe that they are worse than others; on easy tasks, people underestimate their actual performances but believe they are better than others. This paper offers a straightforward theory that can explain these inconsistencies. Overprecision appears to be more persistent than either of the other two types of overconfidence, but its presence reduces the magnitude of both overestimation and overplacement.

Incentives in Experiments: A Theoretical Analysis

Experimental economists currently lack a convention for how to pay subjects in experiments with multiple tasks. We provide a theoretical framework for analyzing this question. Assuming statewise monotonicity and nothing else, we prove that paying for one randomly chosen problem—the random problem selection mechanism—is essentially the only incentive compatible mechanism. Paying for every period is similarly justified when we assume only a “no complementarities at the top” condition. To help experimenters decide which is appropriate for their particular experiment, we discuss empirical tests of these two assumptions.

Prediction Markets: Alternative Mechanisms for Complex Environments with Few Traders

Double auction prediction markets have proven successful in large-scale applications such as elections and sporting events. Consequently, several large corporations have adopted these markets for smaller-scale internal applications where information may be complex and the number of traders is small. Using laboratory experiments, we test the performance of the double auction in complex environments with few traders and compare it to three alternative mechanisms. When information is complex we find that an iterated poll (or Delphi method) outperforms the double auction mechanism. We present five behavioral observations that may explain why the poll performs better in these settings.

Designing stable mechanisms for economic environments

We study the design of mechanisms that implement Lindahl or Walrasian allocations and whose Nash equilibria are dynamically stable for a wide class of adaptive dynamics. We argue that supermodularity is not a desirable stability criterion in this mechanism design context, focusing instead on contractive mechanisms. We provide necessary and sufficient conditions for a mechanism to Nash implement Lindahl or Walrasian allocations, show that these conditions are inconsistent with the contraction property when message spaces are one-dimensional, and then show how to use additional dimensions to achieve dynamic stability while gaining budget balance out of equilibrium.

Generalized Groves–Ledyard mechanisms

Groves and Ledyard (1977) construct a mechanism for public goods procurement that can be viewed as a direct-revelation Groves mechanism in which agents announce a parameter of a quadratic approximation of their true preferences. The mechanisms Nash equilibrium outcomes are efficient. The budget is balanced because Groves mechanisms are balanced for the announced quadratic preferences. Tian (1996) subsequently discovered a richer set of budget-balancing preferences. We replicate the Groves–Ledyard construction using this expanded set of preferences, and uncover a new set of complex mechanisms that generalize the original Groves–Ledyard mechanism. The original mechanism, however, remains the most appealing in terms of both simplicity and stability.