Ph.D. Program in Information and Communication Technologies Curriculum on Electronics and Computer Engineering Dissertation MODELING AND STATISTICAL ANALYSIS OF FINANCIAL MARKETS AND FIRM GROWTH by LINDA PONTA

The quantitative study of financial markets is more and more widespread due to their growing importance in the economy and everyday life. Financial markets can be viewed as real-world complex dynamical systems which are continually evolving, have significant practical importance and produce an enormous amount of data recording the aggregate action of many participants. In recent years, the quantitative approach to the modeling of stock markets has greatly benefitted from methods and tools developed in the domains of engineering and physics. Indeed, the recent availability of large and high quality data sets has transformed finance into the most quantitative social science. A field called “Financial Engineering” has appeared, gathering scientists experienced in non-linear, deterministic and stochastic dynamical systems and interested in modeling and forecasting financial markets. Many world-class US universities have opened laboratories, departments or courses on financial engineering. Moreover, the study by means of simulations of complex systems, characterized by the interaction of a large amount of heterogeneous units, needs one to develop flexible software frameworks based on advanced software engineering techniques. The growing interest in this kind of approach is further confirmed by an emerging field of statistical physics, called “Econophysics”. The first part of this Thesis presents an artificial financial market conceived as a computational experimental facility where different experiments can be performed, hypotheses verified and conjectures validated. An artificial financial market is an agent-based computer simulator of a financial market. Artificial financial markets model financial interactions from the bottom up by means of a large number of interacting agents. This approach relies heavily on computational tools in order to avoid the restrictions of analytical methods. The analysis of financial markets through the construction of artificial markets aims to explain the emergence of the characteristic statistical properties of asset prices on the basis of hypotheses on traders behavior, market microstructure and economic environment. These problems are usually too complex to be treated analytically, thus computer simulations have to be employed. This approach involves the identification of an interactions system that allows the researcher to generate financial time series in order to study the relationships between the elements of the system and market results. The dynamics of a financial market depends on the interactions between the rules defining the trading mechanism and the behavioral assumptions about the agents population. Therefore, building an artificial market means to determine the trading 1See, for instance, the Operations Research and Financial Engineering Department of Princeton University (http://www.orfe.princeton.edu/), the Laboratory for Financial Engineering at MIT (http://lfe.mit.edu/) and the Master degree on Financial Engineering at Berkeley (http://www.haas.berkeley.edu/MFE/).