Measuring Japanese Stock Correlation Networks and Application for Investment Strategy

Financial markets exhibit complex and dynamical correlations between assets, which are generally represented by networks. Such networks are usually constructed from correlation matrices and have been rich subjects to study over past two decades both for academicians and financial practitioners, as these correlation-based networks reveal interconnected structures among assets and characterize market instability, with making strong use of topological features derived from network theory. In this work, we introduce several network measures for graphs, which are applied to Japanese stock market to monitor the market dynamics over a 20-years period. We observe network measures can detect large market events and are consistent with market volatilities. We further examine the relation between network measures and stock returns, by means of which we propose the investment strategy. The backtesting result shows network measures have a potential to mitigate risks and, at some case, the proposed strategy outperforms the benchmark strategy. We finally study the dynamics of network measures and market behavior, which we find are linked to each other.