An economic model is a theoretical construct used by economists to represent economic processes in a simplified way and study their structure. Normally they are based on logical relationships of cause-and-effect between two or more economic variables, then made explicit in mathematical form.
Most economic models apply classical equilibrium theories to represent the relationships between variables. These models are called factor-based models: factors enjoy linear and static relationships, they are based on the assumption that it is sufficient to study collective phenomena at an aggregate level to understand the functioning of the entire system with a top-down approach. One of the strongest assumptions present in these models is the rationality of the agents. Together with the absence of information asymmetries, this assumption generates "frictionless" interactions, and therefore qualitatively different from reality.
As mentioned above, these models assume that there is a long-term equilibrium, but due to non-rational interactions of the agents there can be more or less significant deviations; moreover, the concept of equilibrium, being placed in the long term, is in itself asymptotic. One of the most famous factor-based models is the Capital Asset Pricing Model, which forms the basis of modern portfolio theories.
In order to solve this problem, over time, models that adopt an opposite approach, i.e. bottom-up, have begun to emerge. The so-called agent-based models are a class of computational models developed since the late '40s and are used to simulate interactions between agents in order to study their effects on the economic system. These types of models are useful to analyze complex systems - a concept stolen from physics - i.e. dynamic systems formed by different elements that have non-trivial interactions between them. They are therefore not necessarily linear, placing them already in another dimension with respect to classic factor-based models. The greater the quantity and variety of relationships among the elements of a system, the greater its complexity.
Agent-based models are extremely useful in all those cases in which well-defined decisions and behaviors are observable, with consequent adaptations of the agents. It is therefore assumed their dynamism: in these models the agent is not static, but is confronted with the complexity of the environment and takes decisions accordingly, learning and adapting to the surrounding environment. A classic application is found in financial markets, in an attempt to explain the trend: an example can be found in the causes of recent financial crises and speculative bubbles.
I mercati finanziari rappresentano un caso esemplare di sistema dove le interazioni possono essere estremamente complesse, prestandosi quindi perfettamente per un’analisi agent-based.
Written by Sara Ceccato of the VGen Finance Hub
