Measuring systemic risk of the US banking sector in time-frequency domain

Abstract

To estimate short-term, medium-term, and long-term financial connectedness, we propose a frequency-based approach and measure the contribution of individual financial institutions to overall systemic risk. We derive Wavelet Conditional Value at Risk (WCoVaR) – a robust market-based measure of systemic risk across financial cycles of differing length. We evaluate the systemic importance of financial institutions based on their stock returns and use wavelet framework to analyze returns in a time-frequency domain. Empirical analysis on US banking sector data between 2004 and 2013 demonstrates that wavelet decomposition can improve the forecast power of the CoVaR measure. We use panel regression to explain systemic importance of individual banks, using their objectively measurable characteristics and conclude that size, volatility and value-at-risk are the most robust determinants of systemic risk.