In the agent-based approach, financial
markets are modelled as systems of interacting agents and several examples of such
models have been successful in reproducing the stylized facts that are common to a wide variety of
markets, instruments and periods. By finding economic explanations for the statistical signatures of these market fluctuations, agent-based models can inspire investors and regulators to conceive tools and policies for improved financial decisions and financial risk management. The study of the multi-asset case is also the road to understand the non trivial correlations between assets, an area of research that is receiving recently an intense interest with the use of random matrix theory, complex networks, and multi-scaling.
Correlations among returns of different assets are highly unstable and this is a major challenge for portfolio optimization, for the pricing of multi-asset derivatives and for co-integration based trading
strategies. The Authors approach aims also at bringing insights on the direct impact of trading strategies on speculative markets dynamics compared to the importance given to microstructure effects, e.g in the zero-intelligence agents model studied in Daniels et al and to the importance given to the topology of interactions between agents in Cont and Bouchaud and Iori. The Authors thus extend the agent-based model to the multi-asset case and, in order to obtain transparent pictures, we focus on agents diversifying their strategies among two assets. The article is structured as follows. Section 2 summarizes the single asset model and discusses its main properties. The multi-asset market model described in section 3. The numerical effects of the diversification discussed in section 4. Conclusions are drawn in the last section.