Change point detection for subprime crisis in American banking: From the perspective of risk dependence

Abstract

The subprime crisis has received great attention in academic research but there is no consensus on when the crisis started and when it ended. Previous researchers have only mentioned their subjective judgments in related papers and well-accepted change point detection methods are not available. So the objective of this paper is to propose a multiple change point detection approach from the perspective of risk dependence by using copula function. Since the inter-dependence of different types of risks during crisis and non-crisis periods is significantly different, we monitor the change of dependence structure over time. The first step is to choose a proper copula that can accurately describe the dependence structure of the data. Thereafter, using the chosen copula to fit the data dynamically, a series of parameters are attained. Finally, the change points are identified by analyzing the trend of the fitted parameters. Based on the financial data of the top 100 American banks in Forbes' list, we empirically detect the start point, end point and peak period of the subprime crisis in American banking. The results show that the crisis started in 2007Q4 and ended in 2011Q3, and the peak period of the crisis was from 2009Q3 to 2010Q2.