| This dissertation explores the behavior of economic agents in distributed network environments, such as the Internet. The first chapters are devoted to describing the results of a laboratory experiment on learning in dynamic, real-time, low-information settings. Little evidence of convergence to the Nash equilibrium is found for a dominance solvable version of a Cournot oligopoly. In an asynchronous setting, characterized by players updating strategies at different frequencies, play favors slower learners. Convergence is more robust for a “serial cost sharing” game, which satisfies a stronger solution concept of overwhelmed solvability. As the number of players grows, convergence in the serial cost sharing game tends to diminish. This is driven by a cascading effect in which frequent and highly structured experimentation by one player induces experimentation by others.; In chapter four, we consider the ability of existing learning models to explain the path of play observed in the experiments. Surprisingly little research has focused on how well economic learning models describe human adaptation in dynamic environments. It is shown that history-dependent reinforcement learning models are ill suited to these environments, in which sastisficing models accurately predict behavior. A further objective is to determine which heuristics, or “rules of thumb,” most accurately capture the responsiveness of subjects when incorporated into learning models. Reference points and a particular type of experimentation are found to be important for predicting play.; In chapter five, we analyze a specific issue arising on the Internet—illicit copying of information goods. As new technologies make the reproduction and storage of CDs, movies, and books effectively free, and allow copies to be made without any loss of quality, traditional analysis would suggest that firm profits should diminish as a result of copying. It is demonstrated that this need not be the case. Akin to bundling, copying can be viewed as selling a good to an aggregate consumer, embodying the valuations of a number of agents. Since summing valuations can reduce the dispersion of values, greater surplus extraction may be obtained. |
