Complexity in Financial Markets: Modeling Psychological Behavior in Agent-Based Models and Order Book Models

The fundamental idea developed throughout this work is the introduction of new metrics in Social Sciences (Economics, Finance, opinion dynamics, etc). The concept of metric, that is the concept of measure, is usually neglected by mainstream theories of Economics and Finance. Financial Markets are the natural starting point of such an approach to Social Sciences because a systematic approach can be undertaken and the methods of Physics has shown to be very effective. In fact since a decade there exists a very huge amount of high frequency data from stock exchanges which permit to perform experimental procedures as in Natural Sciences. Financial markets appear as a perfect playground where models can be tested and where repeatability of empirical evidences are well-established features differently from, for instance, Macro-Economy and Micro-Economy. Thus Finance has been the first point of contact for the interdisciplinary application of methods and tools deriving from Physics and it has been also the starting point of this work. We investigated the origin of the so-called Stylized Facts of financial markets (i.e. the statistical properties of financial time series) in the framework of agent-based models. We found that Stylized Facts can be interpreted as a finite size effect in terms of the number of effectively independent agents (i.e. strategy) which results to be a key variable to understand the self-organization of financial markets. As a second issue we focused our attention on the order book dynamics both from a theoretical and a data oriented point of view. We developed a zero intelligence model in order to investigate the role of vanishing liquidity in the price response to incoming orders. Within the framework of this model we have analyzed the effect of the introduction of strategies pointing out that simple strategic behaviors can explain bursts of intermittency and long memory effects. On the other hand we quantitatively showed that there exists a feedback effect in markets called self-fulfilling prophecy which is the mechanism through which technical trading can exist and work. This feature is a very interesting quantitative evidence of a self-reinforcement of agents’ belief. Last but not least nowadays we live in a computerized and networked society where many of our actions leave a digital trace and affect other people’s actions. This has lead to the emergence of a new data-driven research field. In this work we highlighted how non financial data can be used to track financial activity, in detail we investigate query log volumes, i.e. the volumes of searches for a specific query done by users in a search engine, as a proxy for trading volumes and we find that users’ activity on Yahoo! search engine anticipates trading volume by one-two days. Differently from Finance, Economics is far from being an ideal candidate to export the methodology of Natural Sciences because of the lack of empirical data since controlled (and repeatable) experiments are totally artificial while real experiments are almost incontrollable and non repeatable due to a high degree of non stationarity of economical systems. However, the application of method deriving from complexity to the Economics of Growth is one of the more important achievement of the work here developed. The basic idea is to study the network defined by international trade flows and introduce a (non-monetary) metric to measure the complexity and the competitiveness of countries’ productive system. In addition we are able to define a metric for products’ quality which overcomes traditional economic measure for the quality of products given in terms of hours of qualified labour needed to produce a good. The method developed provides some impressive results in predicting economical growth of countries and offers many opportunities of improvements and generalizations