Article Preview
Top1. Introduction
Broadman (2000) considered various features of the Russian economy that have hindered competition, arguing that lack of competition was itself hindering economic growth. In fact, the need for improved competition has long been recognised within the Russian Federation (and earlier, the USSR), especially in the commodity and financial services sectors where dominance of the market is subject to strict anti-monopoly regulation (Zimbler and Okorochenkov, 2006). The Law on the Defence of Competition defines a company to be dominant if it controls more than 50% of market share; collective dominance occurs if any three companies hold more than 50%, or if any five companies hold more than 75%, of market share.
Previous economic modelling has focused on system dynamics and differential equations as a means to model system states with time (Azar, 2012). Agent-based simulations have been used for both intelligent decision making — e.g. in manufacturing (Elghoneimy and Gruver, 2012), real time reasoning (Bevrani et al, 2012) and resource allocation problems (Jiang and Huang, 2012) — and for machine learning of real human behaviours (Hattori et al, 2011). In each case, agent behaviours are determined by the way they respond to signals, both from other agents and from the environment.
In this paper we consider the long-run effects on macroeconomic dynamics of imposing a 50% limit on a single firm’s market share. We take an established equilibrium model of financial flows in the Russian macroeconomy (Ilyasov et al, 2010, 2011a,b), and re-implement it as an agent-based system. This allows us to investigate how the system’s dynamics change with the number of firms participating enabling further experimentation to be done with the model, which an equation-based model does not permit.
We describe the underlying model in more detail in section 3. In section 4 we explain the conversion to a multi-agent system using the state-of-the-art multi-agent modelling platform, FLAME (http://www.flame.ac.uk/), previously used of economic agent-based modelling (Deissenberg et al, 2008; Holcombe et al, 2013). The results are presented in section 5 with a summary and discussion in section 6.
TopAn economy emerges from the interactions of millions of people with firms, banks or governments, all coexisting within the same dynamic environment. Researchers in agent-based computational economics (ACE) use computers to model these actors and their interactions directly as systems of autonomous interacting agents. Allowing the agents to interact over sufficiently long periods leads to various self-organising emergent behaviours which are otherwise hard to identify and formulate in traditional equation-based terms.
Simulating such emergent phenomena traditionally requires the use of powerful supercomputers, but newer techniques, and in particular the use of highly parallel graphics processors, allow basic investigations to be conducted using more modest equipment.