Modelli di Apprendimento e Decisionali in un Mercato Artificiale di Tipo Order-Driven

Progetto: Research project

Description

The research unit's objective is to design an artificial market environment to study how software agents endowed with learning abilities interact and co-evolve overtime. We are particularly interested in investigating the dynamics created by agents with heterogeneous beliefs interacting through an automated trading setting. Wewill model a financial market where several risky assets can be exchanged to analyze the interactions between asset allocation decisions and the dynamics of prices.We will focus our attention on the market volatility generated by the co-evolution of prices and beliefs.The research program will involve the following steps.First, we will model the learning process. We will assume that agents have no structural information about the economy, but they have some beliefs about theevolution of the economy. We will model the diversity of beliefs as the consequence of different interpretations of the available common information. Agents willadjust their beliefs using time series observations on relevant economic variables. In particular, we will allow for the existence of regimes in the joint distribution ofasset returns and we will assume that agents do not observe the state variable, but can use time series data to learn about the state. We will analyze the assetallocation implications of the assumed unobservable regime dynamics.Second, we plan to investigate how the endogenous component of market volatility is affected by the asset allocation dynamics induced by the objective function thatinvestors utilize to determine the optimal portfolio allocations. We plan to compare two cases. One where, as in standard asset pricing models, agents' objective is tomaximize the expected utility of final wealth, and another where we assign to the agents a prospect-type utility function defined in term of upward and downwardmovements of the agent's wealth with respect to a target level of wealth.Third, we will analyse the interactions between the institutional trading mechanism and the dynamics of asset prices. Automated systems may offer advantages interms of operational and trading costs, but they depend on public limit orders for the provision of liquidity. The time variation in liquidity can affect the evolution ofprices, and a complex dynamics may arise between measures of market trading activity and measures of market volatility.Finally, we plan to model the formation and the evolution of groups of agents with correlated beliefs (communication networks). We will assume that agents,cognizant that they possess limited information, understand that alternative interpretations of public information made by other traders may be of value. Under thissetting, agents are not isolated but are connected to other agents (neighbors) and can communicate before updating their beliefs. We plan to structure the agents'population in terms of communities or groups of nodes characterized by having more internal than external connections between them. Using a topological designwith weighted links where the weights are generated dynamically, we will be able to obtain a model where communities emerge endogenously

Layman's description

The research unit's objective is to design an artificial market environment to study how software agents endowed with learning abilities interact and co-evolve overtime. We are particularly interested in investigating the dynamics created by agents with heterogeneous beliefs interacting through an automated trading setting.In the last years, there has been an increasing adoption of automated systems to trade financial securities (See Domowitz and Steil, 1999, for a taxonomy ofautomated systems; and Madhavan, 1992, for an analysis of the major different trading mechanisms). In these settings trading occurs through an electronic orderbook without involving financial intermediaries. Agents must choose, besides the number of units and the price at which to buy or to sell, the type of order to submit inthe market (limit or market order). Automated systems may offer advantages in terms of operational and trading costs, but they depend on public limit orders for theprovision of liquidity. The time variation in liquidity can affect the evolution of prices, and a complex dynamics may arise between measures of market trading activityand measures of market volatility (See Domowitz and El-Gamal, 1999; Coppejans et al., 2001; Domowitz and Wang, 2002; Farmer et al., 2004).Typically the effect of learning on the dynamics of asset prices has been studied assuming either one representative agent learning over time, or a financial marketwhere one single risky asset is exchanged. We plan to overcome both these limitations building an artificial financial market populated by investors withheterogeneous beliefs who can trade several risky assets. Modeling a market with multiple risky securities make it possible to study asset allocation decisions, andtheir effect on the dynamics of prices.The research project will extend the models developed in Consiglio et al., 2005, 2006a, 2006b, and Consiglio and Russino, 2007. In Consiglio et al., 2005, weanalyzed the impact on price changes of the trading mechanism by modeling an economy populated by agents homogeneous in terms of their trading strategy. Eachagent traded to reach an exogenously assigned target portfolio. We showed that the institutional setting of a double-auction market may generate non-normalunivariate marginal distributions of asset returns and temporal patterns resembling those observed in real markets (such as serial dependence in volatility and intrading volume). Moreover, we analyzed the role played by the order submission strategy (that is the decision about the type of order to submit: market or limitorders) specifying a setting where agents choose their order submission strategy on the basis of the information about market conditions revealed by the state of thebook. We showed that the state of the book provides an implicit coordination device inducing agents to supply liquidity when the market needs it, therefore attenuatingthe frictions induced by the trading mechanism. In Consiglio et al., 2006a, 2006b, and in Consiglio and Russino, 2007, we introduced heterogeneous beliefs andendogenous optimal individual portfolio holdings under the assumption of a time-invariant joint distribution of asset returns. We showed that the co-evolution ofinvestor beliefs and market prices generates irregular price series characterized by sharp increases and decreases (looking like bubbles and crashes). Additionally,we provided some preliminary evidence suggesting that the parameters governing the learning process affect market volatility through their impact on the evolution ofmarket liquidity.In this resea
StatoFinito
Data di inizio/fine effettiva9/22/089/22/10