9:00 A Behavioral Model for Mechanism Design: IEL
Jasmina Arifovic, Simon Fraser University
and John Ledyard, California Institute of TechnologyAbstract
We describe a new and different behavioral model for individuals playing in a repeated situation. It is based on a flexible learning process and called Individual Evolutionary Learning. IEL does not require calibration and can be used as a computer testbed to study the probable performance of a wide range of mechanisms over a wide range of environments prior to testing them in a laboratory or using them in practice. We illustrate the utility of the testbed approach by analyzing an open question in mechanism design -- the dynamics of Groves-Ledyard mechanisms. Contrary to standard theories, the prediction from the IEL behavioral model is that the average time to convergence varies smoothly and is U-shaped in the mechanism's free parameter. We validate the results from the testbed with data from economic experiments with human subjects.
Paper here.
10:20 Do stylised facts of order book markets need strategic behaviour?
Dan Ladley and
Klaus Reiner Schenk-Hoppé, University of LeedsAbstract
This paper studies the role of strategy and the order book market mechanism in price dynamics and the order flow behaviour. To this end we analyse a zero-intelligence agent model of a dynamic limit order market. Stylised facts of limit order markets are shown to be influenced and, in some cases, governed by the market mechanism rather than strategic interaction. Positive correlation in order types, for instance, is the result of the market architecture, and price movements may be predicted in the short term from analysing the state of the order book. In contrast the absolute probabilities of order submission highlight the contribution of strategic behaviour.
Paper here.
11:40 The Impact of Heterogeneous Trading Rules on the Limit Order Book and Order Flows
Carl Chiarella, University of Technology, Sydney -- carl.chiarella@uts.edu.au
Giulia Iori, City University -- g.iori@city.ac.uk
and Josep Perelló, Universitat de Barcelona -- josep.perello@ub.eduAbstract
In this paper we develop a model of an order-driven market where traders set bids and asks and post market or limit orders according to exogenously fixed rules. Agents are assumed to have three components to the expectation of future asset returns, namely--fundamentalist, chartist and noise trader. Furthermore, agents differ in the characteristics describing these components, such as time horizon, risk aversion, and the weights given to the various components. The model developed here extends a great deal of earlier literature in that the order submissions of agents are determined by utility maximisation, rather than the mechanical unit order size that is commonly assumed. In this way the order flow is better related to the ongoing evolution of the market. For the given market structure we analyze the impact of the three components of the trading strategies on the statistical properties of prices and order flows and observe that it is the chartist strategy that is mainly responsible of the fat tails and clustering in the artificial price data generated by the model. The paper provides further evidence that large price changes are likely to be generated by the presence of large gaps in the book.
KEYWORDS: Market microstructure, limit orders, fundamentalism, chartism, noise trading, large fluctuations. JEL: C6,D4, G1
Paper here.
1:00 Lunch (suggest: The Blue Stone Cafe on campus, or the AGSM Coffee Shop for sandwiches)
2:00 Wealth Selection in a Financial Market with Heterogeneous Agents
Mikhail Anufriev, University of Amsterdam
and Pietro Dindo, Sant'Anna School of Advanced StudiesAbstract
We study the co-evolution of asset prices and agents' wealth in a financial market populated by an arbitrary number of heterogeneous, boundedly rational investors. Asset demand is proportional to agents' wealth, so that wealth dynamics serves as a selection device. For a general class of investment behaviors, we characterize the long-run market outcome, i.e. the steady-state equilibrium values of asset return, and agents' survival. Our general model illustrates that market forces pose certain limits on the outcome of agents' interactions even within the "wilderness of bounded rationality".
A number of applications are given. We show that our analysis provides a rigorous explanation for the results of the simulation model discussed in the Levy, Levy and Solomon book, The microscopic simulation of financial markets: from investor behavior to market phenomena (2000). We also generalize the model for multi-asset market and explain an observed co-movement of individual stocks, called "market mode".
Paper here.
3:20 Price dynamics with bounded rationality under different market designs
Valentyn Panchenko, UNSW
and Mikhail Anufriev, AmsterdamAbstract
The paper analyzes the dynamics in a model with heterogeneous agents trading in simple markets under different trading protocols. Starting with the analytically tractable model with infinitely large number of agents and Walrasian market clearing, we build a simulation platform to investigate the impact of the trading rules on the agents' ecology and aggregate time series properties. The key behavioral feature of the model is the presence of a finite set of simple beliefs which agents choose each time step according to a fitness measure. The price is determined endogenously and our focus is on the role of the structural assumption about the market architecture.
Paper here.
4:40 Minimalism and Model-Building: An assured model of the exchanges among consumers, retailers, and manufacturers
David Midgley, INSEAD
Robert E. Marks, UNSW
Dinesh Kunchamwar, Barclays Capital
Daniel Klapper, University of FrankfurtAbstract
Previously we presented an agent-based model (ABM) of the exchanges between consumers, retailers and manufacturers. More recently we published a procedure for "assuring" ABMs in general, using this model as an illustration. Model assurance combines ideas from software proof, destructive testing and empirical validation. In that paper we raise the philosophical issue of whether social scientists should take a traditional scientific approach to building ABMs or whether they should prefer a minimalist approach. Taking our own advice we developed a second, minimalist version of our model--Supermarket ABM 2.0. In this paper we present the results of assuring Supermarket ABM 2.0.
The specific steps taken to assure the model include: Verification (Two external experts have inspected the RePast code to discover whether it follows the model specification correctly. We use the Genetic Algorithm as an optimizer to test the bounds of the model by seeking implausible results.) Validation (A real supermarket chain has provided two databases which we use to validate the model. Here we follow a hybrid approach where we use one database to calibrate consumer agents at the micro-level and then we fit the retailer and manufacturer models to the other database at the macro-level, again using the GA.)
We report the results of this model assurance exercise and use it to define "minimalism" more tightly, arguing that it is more restrictive than "parsimony." We also extend this debate by discussing the practical barriers that currently prevent ABMs reaching their full potential in the social sciences. These include the costs of software proof and the lack of data to validate many aspects of the agents.
Talk here.