ae2010 A Symposium in Agent-Based Computational Methods
in Finance and Economics

6th Edition
Conference Series

Accepted Contributions

Markets and Trading

Agent's Minimal Intelligence Calibration for Realistic Market Dynamics

Iryna Veryzhenko, Olivier Brandouy and Philippe Mathieu

Abstract: This paper investigates the question of how much sophisticated in behavior and intelligence artificial traders need to be in order to replicate both qualitative and quantitative stylized facts within a realistic market microstructure. For this purpose, we introduce an agent-based simulation environment with an architecture close to the Euronext-NYSE Stock Exchange. Series of experiments with different kinds of agents’ behavior and trading framework specifications were realized within this environment. The results indicate that only special calibrations provide realistic stylized facts with coherent quantitative levels. We introduce a new type of agents, called in this paper “strongly calibrated agents", with their specific environment design, that provide price dynamics in quantitative and qualitative accordance with real stock market characteristics.

Trading on Marginal Information
Florian Hauser and Bob Kaempff

Abstract: We present an agent-based simulation of a financial market with heterogeneously informed agents based on a model proposed by Schredelseker (2001). By introducing a modified fundamental trading strategy we extend the model and show that this strategy is a superior choice for most agents in the market. The modified fundamental strategy is characterized by giving more weight to the marginal piece of information an agent receives. We show that this protects agents from making joint mistakes with other market participants and suffering from a herding effect. We also observe that informational efficiency of market prices increases when agents adopt the modified trading strategy.

Stylized Facts Study Through a Multi-Agent Based Simulation of an Artificial Stock Market
Zahra Kodia, Lamjed Ben Said and Khaled Ghedira

Abstract: This paper explores the dynamics of stock market from a psychological perspective using a multi-agent simulation model. We study the stock market trading behavior and the interactions between traders. We propose a novel model which includes behavioral and cognitive attitudes of the trader at the micro level and explains their effects on his decision making at the macro level. The proposed simulator is composed of heterogeneous Trader agents with a behavioral cognitive model and the CentralMarket agent matching buying and selling orders. Simulation experiments are being performed to observe stylized facts of the financial times series and to show that the psychological attitudes have many consequences on the stock market dynamics. These experiments show that the modelization of the micro level led us to observe emergent socio-economic phenomena at the macro level.


A Variable bid Increment Algorithm for Reverse English Auction
Imène Brigui-Chtioui and Suzanne Pinson

Abstract: In this paper we propose multicriteria strategies for conducting automated reverse English auctions based on software agents. Reverse auctions gained popularity as a result of the emergence of Internet-based online auction tools. A buyer agent negotiates with several seller agents over a single product. The preference model is based on reference points which represent the desired values and the reservation values over each criterion. To insure process evolution, English auctions design often considers a bid increment that represents the minimal amount that a bidder must improve on the current best bid. Generally, the bid increment is fixed before the beginning of the process and kept invariant during the process. Our aim is to allow adjusting the bid increment as the auction process progresses. We propose an anytime algorithm based on an exponential smoothing method that adapts the bid increment to the auction context.

Co-evolutionary Agents in Combinatorial Sealed-bid Auctions for Spectrum Licenses Markets
Asuncion Mochon, Yago Saez, Jose Luis Gomez-Barroso and Pedro Isasi

Abstract: Allocating scarce resources is a difficult duty governments must face. Furthermore, when participants exhibit complex preference structures (substitutes and complements) this task is even trickier. Combinatorial auctions are a good alternative for solving this problem. In this work we have developed a simulator of a combinatorial first-price sealed-bid auction. The bidding behaviour has been simulated by the application of co-evolutionary dynamics in an agent-based model. This model assumes independent bidders with bounded rationality trying to maximize profits. Finally, the simulations have been tested for two environments that involve the sale of spectrum licenses (digital dividend). These techniques are a helpful tool to support governments taking decisions in the awarding process.
The Effect of Transaction Costs on Artificial Continuous Double Auction Markets
Marta Posada and Cesáreo Hernández

Abstract: Fast price convergence and high allocative market efficiency (close to 100%) are two of the most robust results in Experimental Economics. When human-subjects are replaced by artificial-agents, high allocative market efficiency is also attained even if the artificial agents have zero intelligence, but price convergence depends on the agents’ learning. In this paper we study the sensitivity of Continuous Double Auction performance to the imposition of monetary costs in the market. We find that transactions costs reduce market efficiency. Price convergence results are very different when the monetary cost is imposed on the transaction or on the submissions to buy or to sell. Our agent-based market model confirms and extends previous Experimental Economics market results, and provides new behavioral explanations of the price dynamics.


The Rise and Fall of Trust Networks
Kartik Anand, Prasanna Gai and Matteo Marsili

Abstract: The working of economies relies on trust, with credit markets being a notable example. The evaporation of trust may precipitate the economy from a good to a bad state, with long-lasting and large scale structural changes, witness the 2007/8 global financial crisis. Drawing on insights from the literature on coordination games and network growth, we develop a simple model to clarify how trust breaks down in financial systems. We show how the arrival of bad news about a financial agent can lead others to lose confidence in it and how this, in turn, can spread across the entire system. Our model exhibits hysteresis behavior, suggesting that it takes considerable effort to regain trust once it has been broken, emphasizing the self-reinforcing nature of trust at the systemic level. Although simple, the model provides a plausible account of the credit freeze that followed the global financial crisis of 2007/8.

Simulations on Correlated Behavior and Social Learning
Andrea Blasco and Paolo Pin

Abstract: We consider a population of agents that can choose between two risky technologies: an old one for which they know the expected outcome, and a new one for which they have only a prior. We confront different environments. In the benchmark case agents are isolated and can perform costly experiments to infer the quality of the new technology. In the other cases agents are settled in a network and can observe the outcomes of neighbors. We analyze long–run efficiency of the models. We observe that in expectations the quality of the new technology may be overestimated when there is a network spread of information. This is due to a herding behavior that is efficient only when the new technology is really better than the old one. We also observe that between different network structures there is not a clear dominance.

Technology Shocks and Trade in a Network
Davoud Taghawi-Nejad

Abstract: In this paper we show how business cycles can emerge from the interaction of autonomous agents. We devised an agent-based computational microeconomics model of agents who trade in a network of trading partners. We assume that agents who observe decreased profits change their trading partners. At fixed intervals a new production technology becomes available to a single agent. If an agent introduces a new technology he changes his trading pattern and some of his trade partners can have a decrease in profits. The agents who have lower profits start changing trading partners. The change in the trading network can lead to lower production and decreased profits of other agents. Agents with decreased profits also start changing trade partners. In short, the technology shock triggers a snowball effect of agents changing their trading partners; the GDP decreases. When agents find new trading partners and regain their profits the GDP increases. A business cycle emerges.

Wage Discount vs. Premium for the Network Search in the Labour Market
Gergely Horvat

Abstract: In this paper we deal with the question whether using social contacts or "formal" methods (such as application to advertisements) results in higher expected wage for a worker searching for job. Empirically, this question has produced contradictory evidences. In our model we show that one feature of the arrival process of new offers can be crucial for the relation of expected wages: the correlation between the quality of the job an employee holds and the quality of offers she might hear about. If this correlation is sufficiently high the network search gives higher expected wages. The critical value of this parameter negatively depends on the arrival rate of offers, the fraction of new high wage offers, positively hinges on the job destruction rate and the connectivity of the social network. Further, we show that the Rawlsian welfare is the highest while the utilitarian welfare might be the lowest when the network search gives the highest expected wage. We use mean-field approximation and simulations to obtain our results.


The (Beneficial) Role of Informational Imperfections in Enhancing Organisational Performance

Friederike Wall

Abstract: The paper analyses the effects of imperfect information on organisational performance under the regime of alternative organisational settings. The analysis is based on an agent-based model which is an extended variant of the NK model. In the simulations, fitness measurements are distorted with imperfections according to information asymmetries that are related to differentiation and delegation of decision-making. The results indicate that the effects of informational imperfections on organisational performance subtly interfere with coordination mode, incentives and intra-organisational interactions. The results might throw some new light on imperfect information as in some organisational settings rather insignificant performance losses compared to perfectly informed decision-makers occur, and in some settings imperfect information turn out to be beneficial.

Social Interactions and Innovation: Simulation Based on an Agent-Based Modular Economy
Bin-Tzong Chie and Shu-Heng Chen

Abstract: Using an agent-based modular economic model, we study the effect of social interactions on product innovation and its further impact on competitiveness dynamics. Two firms with different intensities of social interactions are placed in a context of duopolistic competition. The macroscopic analysis based on various criteria, including the market share, profit rate, accumulated capital, waste ratio and consumers’ satisfaction level, indicates that high social interaction within the firm can lead to not only a healthy firm but also a healthy economy. However, this positive result is undermined by the catastrophic nature of the modular economy as shown in the microscopic analysis. Furthermore, the mesoscopic analysis shows that in the long run the duopoly market tends to become a monopolistic market, and there is a non-trivial probability that the low-interaction firm will drive out the high-interaction firm. The risk of innovation in this model may be greater than what the usual economic model may expect.

Threshold Rule and Scaling Behavior in a Multi-Agent Supply Chain
Valerio Lacagnina and Davide Provenzano

Abstract: In this paper an agent-based model of self organized criticality is devel- oped in a network economy characterized by lead time and a threshold behavior of firms. Instead of considering the aggregate production of the economy as a whole, we focus on both the propagation and amplification effects of a demand shock in the sectorial productions of a multi-agent supply chain. We study a static network structure representing a relation of firms in a lower-upper stream in an industrial organization. In our model, the individual (R, nQ) policies play an important role in generating a propagation effect across the different layers of the economy, and the propagation turns into the large fluctuations and amplifications of sectorial productions.

Industry Sectors

Information and Search on the Housing Market: An Agent-Based Model
John Mc Breen, Florence Goffette-Nagot and Pablo Jensen

Abstract: We simulate a closed rental housing market with search and matching frictions, in which both landlord and tenant agents are imperfectly informed of the characteristics of the market. Landlords decide what rent to post based on the ex- pected effect of the rent on the time-on-the-market (TOM) required to find a tenant. Each tenant observes his idiosyncratic preference for a random offer and decides whether to accept the offer or continue searching, based on their imperfect knowledge on the distribution of offered rents. The steady state to which the simulation evolves shows price dispersion, nonzero search times and vacancies. We further as- sess the effects of altering the level of information for landlords. Landlords are better off when they have less information. In that case they underestimate the TOM and so the steady-state of the market moves to higher rents. However, when landlords with different levels of information are present on the market, the better informed are consistently better off. The model setup allows the analysis of market dynamics. It is observed that dynamic shocks to the discount rate can provoke overshoots in rent adjustments due in part to landlords use of outdated information in their rent posting decision.

Adaptation of Investments in a Pharmaceutical Industry
Tino Schüette

Abstract: Situated in the research field of market structure and strategic behavior, a model of product market competition is developed, showing the impacts of invest- ment adjustments on the success of companies. Placed in an agent-based multi-firm multi-product setting, we study the consequences of the two-step decision of firms to split their budgets in (i) marketing and development activities and (ii) develop- ment expenditures into innovative or imitative activities. The model is validated with empirical data of the pharmaceutical industry, with reference to the drug market in Germany. The results show that investment strategies adjusted to the behavior of direct competitors outperforms adjustments based on individual aspiration levels.
An Agent-Based Information Management Model of the chinese Pig Sector
Sjoukje A. Osinga, Mark R. Kramer, Gert Jan Hofstede, Omid Roozmand and Adrie J.M. Beulens

Abstract: This paper investigates the effect of a selected top-down measure (what-if scenario) on actual agent behaviour and total system behaviour by means of an agent-based simulation model, when agents’ behaviour cannot fully be managed because the agents are autonomous. The Chinese pork sector serves as case. A multilevel perspective is adopted: the top-down information management measures for improving pork quality, the variation in individual farmer behaviour, and the interaction structures with supply chain partners, governmental representatives and peer farmers. To improve quality, farmers need information, which they can obtain from peers, suppliers and government. Satisfaction or dissatisfaction with their personal situation initiates change of behaviour. Aspects of personality and culture affect the agents’ evaluations, decisions and actions. Results indicate that both incentive (demand) and the possibility to move (quality level within reach) on farmer level are requirements for an increase of total system quality. A more informative governmental representative enhances this effect.

Wealth Distribution Evolution in an Agent-Based Computational Economy
Victor Romanov, Dmitry Yakovlev, and Anna Lelchuk

Abstract: In this paper we study the modification of wealth distribution among the customers during quite a long period of time in the model — several model years. During this time customers get their income in forms of salary depending on enterprise production volume and assortment, or redundancy payments. As a part of the study it was detected that whilst the initial wealth distribution was uniform a strong non-uniformity arises after several years in the model. The model includes the following interacting agent classes: customer, bank, labor market, state, enterprise, market, university, and mass media. The model also allows us to evaluate the relations among the efficiency of enterprises’ investment strategies, tax level and customer’s prosperity and unemployment level. The possibility of obtaining a new specialty by a fired agent for the purpose of stabilization and increasing his profit and improve standard of life is considered in the paper as well.

Endogenous Credit Dynamics as Source of Business Cycles in the EURACE Model
Andrea Teglio, Marco Raberto and Silvano Cincotti

Abstract: The paper investigates the relationship between the amount of credit money in the economy and the variability of output and prices in the EURACE model. First we examine if the decision about dividends payment by the firms can affect this variability, then we adopt the policy measure of quantitative easing, that has been largely used by the Fed and the Bank of England during the recent crisis, in order to understand its effect on economic instability. Results show the emergence of endogenous business cycles which are mainly due to the interplay between the real economic activity and its financing through the credit market. In particular, the amplitude of the business cycles strongly raises when the fraction of earnings paid out by firms as dividends is higher, that is when firms are more constrained to borrow credit money to fund their activity.
Reinforcement Learning of Heterogeneous Private Agents in a Macroeconomic Policy Game
Mahdi Hemmati, Masoud Nili and Nasser Sadati

Abstract: A repeated inflation-unemployment game within the linear-quadratic framework of Barro and Gordon is studied assuming that the government would like to cheat optimally and the finite heterogeneous population of private agents attempts to learn the government’s targets using a reinforcement learning algorithm. Private agents are heterogeneous in their initial expectations of inflation rate but are assumed to utilize an identical anticipatory reinforcement learning process, namely Q-learning. In our heterogeneous setting, the only way for the private agents to achieve a zero value for their loss function, is for all of them to correctly anticipate the Nash equilibrium. It is of particular significance that such a solution requires a convergence of expectations across an initially heterogeneous population. Computer simulations have been conducted using different tuning parameters to investigate the convergence of our proposed model of learning process to Nash equilibrium.

Demographics and Culture

Towards and Agent-Based Model of the Economic Development Process: The Dynamics of the Fertility Rate
Gianfranco Giulioni and Edgardo Bucciarelli

Abstract: This paper is a first step to build an agent-based model of the economic development process. We focus on households’ behavior by studying in particular the relationship between the available income and the optimal choice on quantity (fertility) and quality (level of education) of children. A collection of households taking decisions according to the rules identified at the individual level, but perturbed by idiosyncratic shocks and subject to a mean field interaction are monitored by using computer simulations. The model gives us the opportunity to investigate the evolution of the distributions of fertility and income by using data recorded from simulations at the individual level. Averaging the number of children across households we find that this model to be able to replicate the J-shaped pattern of the fertility rate found in recent empirical analysis.

An Agent-Supported Simulation of Labour and Financial Markets for Migration Processes
Nancy Ruiz, Vicente Botti, Adriana Giret, Vicente Julian, Oscar Alvarado, Victor Perez and Rosa Rodriguez

Abstract: The Migration Process is a phenomenon that includes a variety of actors, societies and political issues at different levels. In the migration problem it is then possible to observe complex interactions among different entities: there are links among economics, politics, social, commercial, labour, health, culture, and safety areas. Migration movements may also influence and be influenced by the effects of policies and norms of both sending and receiving countries. One of the key factors that influence the Migration Process behavior is the Labour Market, which is simultaneously affected by Financial Markets. These interactions have been traditionally represented by mathematical approaches that do not allow including flexibility, autonomy, adaptive and pro-activity features that are present into the dynamic and complex real life migration scenarios. On the other hand, the Multiagent System (MAS) paradigm has been successfully applied in studies related to mass movement in complex environments. In this paper a MAS simulation approach is proposed to simulate the migration process and to model micro-level interaction protocols that link Labour and Financial Markets to Migration Processes (MP-LM&FM) in order to observe dynamic behaviours that may emerge at macro level.
Sensitivity Analysis of an Agent-Based Model of Culture's Consequences for Trade
Saskia L.G.E. Burgers, Gert Jan Hofstede, Catholijn M. Jonker, and Tim Verwaart

Abstract: This paper describes the analysis of an agent-based model’s sensitivity to changes in parameters that describe the agents’ cultural background, relational parameters, and parameters of the decision functions. As agent-based models may be very sensitive to small changes in parameter values, it is of the essence to know for which changes the model is most sensitive. A long-standing metamodeling-based approach of sensitivity analysis is applied to the agent-based model. The analysis is differentiated for homogeneous and heterogeneous agent populations. Intrinsic stochastic effects of the agent-based model are taken into account. The paper describes how an appropriate regression model has been selected and analyses the parameter’s variance contributions in general and in specific cultural settings.