Information Aggregation and Allocative Efficiency in Complex Environments

Citation

Brunner, Christoph (2010) Information Aggregation and Allocative Efficiency in Complex Environments. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/79W2-DW28. https://resolver.caltech.edu/CaltechTHESIS:06092010-233303171

Abstract

It has been suggested that information cascades might occasionally prevent asset markets from performing efficiently. We run experiments in which private signals about an asset with a common value are released sequentially. That allows us to compare the quality of information aggregation in periods in which an information cascade would occur in the absence of prices to the quality of information aggregation in other periods but we find no significant differences.

In a second chapter, we focus on information cascades in sequential games, where subjects choose between two options and each subject has a small chance of being perfectly informed about which option is correct. When subjects can only observe the number of times each option has been chosen by their predecessors, the Nash prediction is that subjects follow the minority of predecessors but subjects tend to follow the majority instead. Models that combine heterogeneity in the level of strategic thinking and allow for some degree of trembling (e.g., noisy introspection) fit the data best.

A third chapter evaluates the performance of four different auction formats. We find that bidders are not always bidding on the currently most-profitable combination of available items. Instead, subjects sometimes submit jumpbids. As a result, a clock auction in which prices can only increase incrementally generates particularly high revenues. We also find that subjects are reluctant to risk exposure. In the presence of value complementarities, introducing the possibility to submit package bids therefore increases efficiency.

In the last chapter, we compare five different stationary concepts: Nash equilibrium, quantal-response equilibrium, action-sampling equilibrium, payoff-sampling equilibrium and impulse-balance equilibrium. Selten and Chmura (2008) run a large number of completely mixed 2 x 2 games in the laboratory for that purpose. We reanalyze their data and find that there are no significant differences with respect to goodness of fit except that the Nash equilibrium fits worse than all the other models. We also estimate the same models on other data sets.

Thesis Files