Simulazione della microstruttura di un mercato finanziario di tipo “order driven” con ottimizzazione di portafoglio

Progetto: Research project

Dettagli progetto


In this project we will explore the interactions between differentmarkets microstructure models and the behavioural assumptions ofartificial heterogenous agents.The motivation to adopt a simulation environment to study suchtopics is concealed in the high complexity of the financialmarkets. Many authors have pointed out that a blend of analyticalmodels and simulation analysis can shed light on most of thepuzzles "echoing" in the financial literature.We are interested in analysing the data series yielded by themarket simulator in order to test their adherence to observedstylized facts. For this matter we need to select a set of testswhich will deem an artificial market as reliable to represent areal financial one. We will also investigate the behaviour ofthe agent-based markets in critical situations, such as, crashesand bubbles. The latter will be a distinguished aspect of ourresearch since financial institutions are compelled to test theirportfolios using stress scenarios, and artificial markets could bethe environment to set up controlled experiments.The approach we propose is a multidisciplinary one, whereresearchers from the cognitive sciences can contribute to shapethe mathematical models used by the economists to explain observedphenomena, and computer scientist, trough efficient simulationtools, can provide a substantial help when the class ofmathematics is too restrictive for the complexity of the realproblems.However, we believe that a robust and meaningful simulationenvironment must be accompanied by a common language among theresearchers of different fields. For this reason we will developan open-source simulation framework which can be easily used anddistributed, and whose aim is to disseminate information in thescientific community and attract practioner of the financialfield.

Layman's description

In a recent essay which retraces the astronomical theory on tides,the science historian L. Russo (2003) proposes an interestingphilosophical cue which will be used throughout this text tomotivate and describe the aims of our project. L. Russo writes:"The description and prediction of tidal waves provide the firstexample of a vast class of problems which can be dealt with tworadically different strategies: the first one is rooted in theancient times and consists of using a simplified model of thephenomenon to be investigated; the second one is quite modern andemploys computer simulation to extrapolate forecasts. In the firstcase we have to renounce to a detailed description of thephenomena under exam; in the second case we have to give up asyntectic picture of the whole. Both relinquishments are too heavyto bear and it is therefore compelling to consider the twostrategies as complementary to each other instead of alternative"The development of Economics as science has witnessed the use ofmathematics as a rigours tool to assess the intuition of thegreatest economists of this century (Walras, Debreu, Samuelson,Keynes, etc.). However, the process of "proofs and refutations"has highlighted that the class of mathematics they were using wastoo restrictive for the class of problem they were dealing with.Indeed the mathematics they dealt with did not work in the realworld, possibly because we needed a richer class of processes.This falsification process has generated an increasing interestand application in economics of ideas from complex systemstheory. The goal of the so-called complexity theorists is to studythe aggregate properties arising in an economical systemrepresented by a set of heterogenous agents whose behaviours areinterdependent.Agent-based simulation models are a third way between fullyflexible but not computable and hardly testable literary models(which provide no more than a verbal description of the causalrelationships behind a given phenomenon) on one side, and moretransparent but highly simplified analytical models.The approach we propose is strongly multidisciplinary. As a matterof fact, it is in this context that the empirical analysis, thecognitive models and the simulation models can provide an addedvalue which is complementary, and not alternative, to themathematical models used to describe the economical phenomena.Under this view, the first objective of our research group is tostudy simulation models of the price dynamics of financialsecurities, starting from the fundamental units which made up themarket, namely, virtual entities which interact each others andwith the institutional structure of the market: call auction orcontinuous trading, order-driven or quote-driven, screen-based orfloor. The flexibility of the simulation models allow differentassumptions on the investor' behaviour and on their riskperception: prospect utility functions, myopic risk aversion,fundamentalist vs. trend chaser. Putting altogether, it will bepossible to analyse the impact of the market structure and of thebehavioral assumptions on what are considered puzzles of thefinancial markets: inefficiency, speculative bubbles and crashes.The second objective of our research group is aimed to establish avalidation "protocol" in order to asses whether a given artificialmarket is reliable to represent a real financial counterpart. Inparticular, the protocol should allow (i) the interpretation ofthe simulation dynamics, (ii) the estimation of the simulationmodel, and (iii) the generalisation of the results.The validation protocol is aimed to the iden
Data di inizio/fine effettiva11/30/0411/30/06