The Reconstruction of Financial Signals Using Fast ICA for Systemic Risk

Independent component analysis (ICA) is a statistical method for transforming multidimensional observed signals into components, which are statistically independent from each other, which is a case of redundancy reduction. In this paper, we implement Fast ICA proposed by Hyvarinen and Oja to investigate the relationship between systemic risk and ICA in the US financial market. We propose a systemic risk indicator based on observing the redundancy level of signals in running 10 variables including 10 S&P 500 sector indices. We find that not only the redundancy level of signals becomes larger during a crisis than during a normal period, but also the financial system becomes more vulnerable when the redundancy level grows up.