Project Archives

Sort by: Title Date Grant

Networked Markets // 2004 - 2010 // SBR-0351801

A central theoretical proposition in economics is that frictionless markets perform at 100% efficiency, and require only trivial links between buyers and sellers. Actual markets, however, face moderate to severe trading frictions, and rely on various sorts of buyer-seller links. We introduce two important frictions: possible cheating (e.g., a seller might ship an item of lower quality) and hidden trading opportunities (e.g., a buyer might not be aware of a low cost supplier). We will then introduce networks of links among traders, and examine the impact on market performance, including efficiency, prices, volume, and profit distribution. As suggested by ratings in on-line auctions or by the role of immigrant networks in international trade, the links among network members spread information, e.g., the identity of cheaters or the location of low cost sellers. Such links may boost market performance by fostering trust (based on reputation) and unifying trade. We will choose the network links in early series of experiments and allow traders to build their own networks in later series.

Landscape Dynamics // 2005 - 2009 // SES-0436509

Supported by NSF grant SES-0436509, this project demonstrates new techniques for social science models. The techniques include nonlinear partial differential equations and agent-based models. The primary application is to financial market dynamics (especially bubbles and crashes), with secondary applications including Downsian voting models and Veblen consumption dynamics. 'Adaptive landscape' is a classic metaphor in theoretical biology. The landscape is defined over an abstract space of potential biological traits, and successful species are represented by peaks, i.e., trait combinations that have the highest fitness. We develop the landscape idea mathematically, drawing on insights and techniques from dynamical systems theory, agent-based systems, evolutionary game theory, and fluid dynamics. The abstract space represents possible choices (in one or more dimensions) by individuals, and the social state is the distribution of actual choices. Then an individual's payoff is a landscape whose shape depends on the social state. Individuals respond to the landscape and seek higher payoff. As they do so, the distribution of choices changes in continuous time, and therefore the landscape also changes. New peaks representing high payoff can appear, and old peaks can disappear, or flatten into plateaus, or move. Thus an application produces a dynamic interplay between social state and landscape. Main website: http://www.vismath.org/research/landscapedyn

Social Preferences & Revealed Altruism // 2002 - 2007 // None

We introduce parametric and non-parametric models of other-regarding preferences, over my own payoff and others' payoffs. My preferences are systematically affected by my emotional state, which responds to relative status and to the kindness or unkindness of others' choices (``intentions''). The emotional state determines the marginal rate of substitution between my own and others' payoffs, and thus my subsequent choices. The models are applied to a variety of existing and new data sets. The Revealed Altruism paper is a copy of the Econometrica article. The Tractable Model paper is the penultimate version of the Games and Economic Behavior article.

Cyberspace Markets // 2000 - 2005 // IIS 9986651

Supported by NSF grant IIS 9986651 (2000-2003). Researched effective cyberspace markets. Measured effectiveness in terms of a market's ability to lead intelligent (but self-interested) traders quickly to efficient transactions and to encourage their participation. Methodologies: * theoretical models of market equilibrium and transient performance; * computer simulations of traders and their market interactions; * laboratory markets with human traders interacting with automated traders; and * statistical analysis of recent cyberspace activity.

Local Interactions // 1997 - 2001 // None

Latest works in economics have focused on local communities of interactions as a way to model sustainable cooperation - in Prisoner's Dilemma games - and relatively fast coordination on the risk-dominant equilibrium - in Coordination games. We tested in the laboratory some new network theories on how does the network architecture - the pattern of links between individuals - encourage coordination and cooperation. We implemented a computational model - in Swarm - to study how the pattern of relations linking financial institutions - the network - affects the diffusion of financial crises.

Customer Markets // 1997 - 2001 // SBR 9617917

Markets with switch costs and consequent buyer-seller attachments are called customer markets, following Okun (1981). Customer markets are interesting for three complementary reasons. First, they are pervasive in modern economics, with a major share of wholesale and intermediate transactions as well as labor and retail. Second, their performance characteristics may differ from auction markets. Third, they are not yet well understood, either theoretically or empirically. Supported by NSF grant SBR 9617917 (1997-2000) we use laboratory experiments to expand empirical knowledge of customer markets.

Optimal and Adaptive Learning Models // 1993 - 1997 // SBR-9310347

This grant funded a series of laboratory experiments of individual choice under uncertainty. The primary task, called medical diagnosis, presented the subject with binary-valued cues ("patient's symptoms") and asked for a binary or continuous response representing the more likely disease (or its probability). The results are reported in five academic papers: "Learning to Forecast Price," "Understanding Variability in Binary and Continuous Choice," "Monty Hall's Three Doors: Construction and Deconstruction of a Choice Anomaly," "Broadening the Tests of Learning Models," "A Comparison of Learning Models," We also report results in: Hugh M. Kelley and Daniel Friedman, "Learning to Forecast Rationally," in The Handbook of Experimental Economic Results, C. Plott and V. Smith (eds.) (2000), Elsevier Press, Amsterdam, forthcoming.

Price Formation and Learning // 1993 - 1996 // SBR-9223461

A series of laboratory experiments that examine price formation in double auction markets, with a focus on the dynamical processes which may lead to market equilibrium. The experiments feature random individual value and cost parameters independently drawn from known uniform distributions in each trading period.

Evolutionary Game Experiments // 1991 - 1993 // SES-9023945

An empirical study of the process by which Nash equilibrium (or other coherent behavior) may be achieved by intelligent self-interested agents in strategic interaction. We present populations of 6-24 undergraduate subjects with chosen payoff (or fitness) functions which embody interesting sorts of strategic interaction. Each subject faces a given payoff function and interacts anonymously with a fixed population of other subjects over 10-100 decision periods. Analysis compares actual time paths of population distributions to various theoretical time paths proposed by biologists and economists, including constant Nash equilibrium and refinements of Nash equilibrium.

Financial Market Mechanisms // 1988 - 1991 // IRI-8812789

Examined alternative electronic trading formats for asset markets, including clearinghouse (or call) and continuous double auction. Focused on the efficiency impact and allocation of surplus when different traders had different privileges, e.g., ability to post price offers. Anticipated (by too many years!) and tried to inform the move to electronic crossing networks, which now carry a large fraction of asset trading volume in the field.

Experimental Asset Markets // 1984 - 1987 // SES-8411382

Starting at UCLA in 1984, and continuing at UCSC in 1985, we constructed the world's first all-electronic asset trading platform. Traders received public and private information during trading periods. Lead programmers Hekke Ketola at UCLA and April Atwood at UCSC implemented the UNIX platform, which ran on a server in the PI's office over the campus network, and subjects used workstations in campus teaching labs.

Current