|
|
Reactive Multi-agent models in economics
Burian, Jan ; Pstružina, Karel (advisor) ; Berka, Petr (referee) ; Pospíchal, Jiří (referee)
Multi-agent models are software models of real systems, which are compounded of a large amount of active autonomous units -- agents. The main theme of this study is the application of multi-agent models in economics. In the first part of this study I synthesized information from many sources and created a proposal of methodology of multi-agent modeling in social sciences. This methodology is illustrated by a detailed analysis of a well-known model of ethnocentric behavior. I focused on the issue of complexity a self-organization a observed the development of different kinds of structural complexity. Another important issue I'm dealing with is the relation between multi-agent models and analytical models used by the mainstream economy The second part contains three applications of multi-agent models in economy. All described models are research models -- the aim is to get insight into the fundamental laws controlling the dynamics of the system. Research models should be as simple as possible -- than we can understand how the behavior of the system is resulting from the interaction of agents. All models represent an alternative to some typical economic situation. The alternative consists mainly in abandoning of the assumption of full rationality of agents. I focused on models with reactive agents, which don't dispose any memory (or only an indirect form of memory) and have no ability of complex symbolical representation of the environment. Reactivity of agents doesn't necessarily mean an unwanted limitation. It could be properly used to model different kinds of bounded rationality. The model of transaction costs on financial markets is based on another model, which showed how the instability of the financial market could result from interactions of agents with bounded rationality. The instability of the financial markets contradicts the Effective Markets Hypothesis. I modified and extended this model with the mechanism of transaction cost. The model shows that higher transaction costs lead to greater stability of the market, they decrease the volume of trade and to some extent they can increase the efficiency of the modeled market. The model of evolution of aversion to risk and ambiguity is inspired by psychological and neurological experiments whose results contradict the Expected Utility Hypothesis. Model shows that there exists very simple mechanism which leads to evolution of risk and ambiguity aversion. This aversion evolves because of situations when the agent can repeatedly decide between two choices, one brings him low but risk free utility, second brings him higher utility which involves the risk that the agent will lose its ability to get any utility in the future (e.g. it will die). The model shows that sexual reproduction strengthens this mechanism. If this mechanism has also evolved in humans, it can influence their behavior and lead to "irrational" behavior contradicting the Expected utility hypothesis. The model of monopsony on the labor market analyses the situation in which the employer could cut the wages to the existential minimum of workers. In such a case are the workers paradoxically to increase the labor supply when the wages are decreasing. The model further analyses the influence of consumer loans on the profit of the employer, unemployment and welfare. The model shows that the increasing of the probability of consumer loans increases the profits, increases unemployment, decreases well-fare and in some situations it enables the employer to decrease the wages to the existential minimum. The model also suggests that for very low wages the decrease of minimum wage could lead to increase of unemployment. From the results recommendations could be obtained both for the profit maximizing employer and for the workers resisting the wage cuts.
|
guest ::
