2001 Graduate Workshop in Computational Economics

Student Projects

Each student began a research project during the two-week workshop. Below are brief descriptions of these various projects. These projects will form the basis for dissertation chapters and/or journal articles.

Alessandro Acquisti, SIMS, UC Berkeley (acquisti@sims.berkeley.edu).

Alessandro is studying how shared knowledge influences production of complex goods. Such goods are ubiquitous in computational systems, where the interacting knowledge of creators, mediated through computers and communication networks, creates new goods. Alessandro uses an NK-landscape to model the system. Each agent faces a binary search problem, the solution of which depends both upon the agent's choices and the choices of the other agents. As the network grows the agents are able to better incorporate the activities of the other agents in their search. He finds that network effects can dramatically alter performance and the effectiveness of various learning strategies.

Keyvan Amir-Atefi, Economics, UC Santa Barbara (keyvan@econ.ucsb.edu).

Keyvan is exploring the performance of trading agents in multiple markets. The agents are drawn from the current literature in economics, including variants of informed and uninformed traders, and their actions are synthesized by a Bayesian market specialist. Over time, more successful agents are increased via a replicator dynamic. He finds that the dynamics of the model can lead to dramatic changes in the price as destructive feedback loops form when the informed agents over react to price signals and agglomerate on particular markets.

Karin Andersson, Engineering Physics, Chalmers University of Technology (f95kaan@dd.chalmers.se).

Karin is looking at buyer and seller behavior in auctions of bundled goods. Her initial work focuses on the study of dynamics in an auction with a single seller and adaptive buyers. This work is then extended to an auction in which the seller analyzes all of the incoming bids to better identify and exploit the latent demand of buyers. She finds that buyers can learn to productively bid in the simple version of the game, though the ability to do so is closely tied to the learning mechanism. The work is currently being extended to the full coevolutionary model.

Chris Brooks, Computer Science, University of Michigan (chbrooks@umich.edu).

Chris is looking at niche formation in information economies. In the model producers attempt to attract consumers by offering different bundles of information goods. Consumers pick the best bundle from those offered by the producers, based on their valuations of each bundle's attributes tempered by the ``clutter cost'' of having to read things that are not of interest. Producers use a genetic algorithm to search for profitable bundles. He finds that periods of niche consolidation by the producers alternate with periods where one producer or the other dominates. If producers are allowed to coordinate their searches, they can both achieve higher profits as long as the coordination is limited. Moreover, small increases in clutter cost improve the ability of firms to make higher profits.

Jung-Kyoo Choi, Economics, University of Massachusetts at Amherst (jungk@econs.umass.edu).

Jung-Kyoo is analyzing the evolution of cooperation in N-person social dilemmas. Agents are assigned into groups and play a repeated N-person social dilemma. Agents either interact with a random group (global) or the same group (local). A genetic algorithm is used to adapt the strategies with a selection mechanism that involves either global or local agents. With random groupings, defection is the norm with some occasional, short-lived outbreaks of cooperation (these outbreaks are enhanced when selection is local). When agents stay within the same group cooperative outcomes are much more likely, with relatively high levels of sustained cooperation in many of the resulting groups when selection is global, but not when it is local.

Ivanna Ferdinandova, Economics, Universitat Autonoma de Barcelona (iferdin@idea.uab.es).

Ivanna is investigating the dynamics of a two-sided adaptive model of buyers and sellers. In the model, coffee sellers decide on what quality of coffee to offer customers based on attempts to adapt policies that lead to higher profits. Buyers have aspirations for coffee quality that adjust based on recent experience, and they search for shops that can meet these expectations. She finds that the willingness of agents to adjust their aspirations to recent consumption has a strong influence on the system's dynamics. When agents adjust aspirations slowly a monopoly forms in the industry. Markets composed of heterogeneous agents result in a single, high-quality firm that shares the market with many, less-profitable competitors. When aspirations are rapidly adjusted a relatively stable market composed of numerous firms ensues.

Michael Heaney, Political Science, University of Chicago (mtheaney@uchicago.edu). Michael considers how learning impacts the development of public policy. The model allows ``states'' that must make a binary policy choice to emulate the choices of other states. Using the model he is developing hypotheses about how policy salience, emulation, and uncertainty interact to produce innovative policies. He is extending the model by incorporating more elaborate learning structures and the use of existing data sets to empirically ground the model.

Maggie Penn, Social Science, California Institute of Technology (epenn@hss.caltech.edu).

Maggie is analyzing city formation. Citizens have preferences over both a private and a public good. Demand for the public good depends on the tax share that the individual must pay, and thus it is linked to the composition of citizens within a given jurisdiction. During each round of the model a new jurisdictional split is proposed and if a majority of the citizens approve the split occurs. Individual citizens are allowed to move if they can find another citizen with whom to swap locations. She finds that the number of cities rises as the square root of the number of agents. She also notes that heterogeneous preferences do not result in dramatically different cities and that the social benefits of new city formation appear to be quite limited in the current model.

Alexander Peterhansl, Economics, Columbia University (ap11@columbia.edu).

Alex is modeling the formation of networks among heterogeneous, game-playing agents. Agents playing a Prisoner's Dilemma game ``grow'' a network of partners with whom to play. Agents use a generalized version of the Tit-For-Tat strategy to play the game within their existing network. Networks evolve as agents, depending on type, either make new connections or incorporate new partners from the sidelines. Similarly, network connections are severed with uncooperative partners. Such partners are returned to the sidelines if they become completely isolated. He finds that inherently stable networks can form in the system and he is currently exploring key hypotheses about the dynamics of the system.

Ritirupa Samanta, Economics, Brandeis University (rsamanta@brandeis.edu)

Riti is trying to understand the dynamics of bank runs and contagion. Both informed and uninformed agents deposit money at multiple banks. Agents withdraw their deposits either when they they need it for private purposes or when they lack confidence in the bank's ability to survive. She finds that certain parameters, such as bank size differences and reserve requirements, can dramatically alter the potential for bank runs and contagion. Based on the model, she is exploring key policy variables that might enhance the stability of banking systems.

John H. Miller , miller@zia.hss.cmu.edu.