Detecting early warning signals of financial crisis in spatial endogenous credit model using patch-size distribution

Abstract

Like numerous complex dynamical systems, financial systems generate abrupt transitions which can cause financial crisis. Since the state of the system generally exhibits little change before such transitions, the sudden transitions are difficult to anticipate. There are considerable literatures for the construction of early warning signals based on features of time series, but there is essential lack of investigating effects of early warning signals at spatial structure for the financial systems. This paper aims at investigating the spatially extended endogenous credit system with stochasticity using spatial early warning signals. First, we apply patch-size distribution to predict upcoming critical transition. The results show that the patch-size distribution conforms a power law at threshold. Second, we also predict impending critical transition utilizing the spatial early-warning indicators widely applied on dynamical systems. These indicators predict successfully the critical transition between high and low asset price state. To enhance the robustness of early warning, we use these methods together to detect impending transition. We hope that our methods can promote the utilization and test of spatial early warning signals on real financial systems.