Spatial aspects of agent-based modeling of large economy

The paper presents a pilot version of spatial agent-based model simulating the national economy of Russia. The model is supposed to be used for evaluation the effects of industrial and spatial policies. The object of modeling is multi-regional economic space of Russia in its interaction with foreign market. The space is physical; locations of agents are defined by geographical coordinates. The basic hypothesis is that decisions of agents on microeconomic level lead to spatial changes on macroeconomic level. We consider Arrow-Debreu model with Leontief technologies as a microeconomic prototype of our model. Theoretical construction of the model, referred to models of stochastically converging equilibrium. The basic features of the model presented are the following. We explicitly account for space, consideringr interactions between trading agents located in physical space. Moreover, the model is compatible and exchanges information with an existent multiregional input-output model that uses the identical. There are agents of the following types: firms, households, commodity markets, labour market and foreign markets. Agents take decisions on the base of microeconomic models under bounded rationality with the account of transportation costs. Geographical pattern of the model is referred to the stylized map of Russia. Distance is measured as a length of the shortest arc between 2 points on the Earth with their coordinates of longitude and latitude. Initialization of the model uses some elements of geo-informational approach as well as informational exchange with multiregional input-output model. Agents are located on the map on the base of geographical coordinates of cities, real data on demography and statistical performance of economic activities. . Statistical performance is collected mostly across macro-regions and activities and is visualized by graphs. The model is realized on "Lua" programming language. Spatial pattern of interactions among agents is created by two-part transport tariff: the first part is proportional to the amount of commodities and the second part is proportional to the distance between 2 agents and to the amount of deliveries. In the report we illustrate the algorithm of trade between households with the account of transportation costs. The geographic structure of commodity flows is analyzed. The report also presents some results of the experiments accomplished. We studied the influence of transportation costs on the convergence of prices to the state of quasi-equilibrium. Experiments demonstrated that the Law of One-Price is not observed. Prices converge to clusters within regions, which reflect the impact of differences in industrial pattern of regions. The prospects of future development of the model are discussed.