Booms and Busts in Asset Prices: Risk Modeling, Bubble Detection, and the Role of Monetary Policy ; Boom-und-Bust-Zyklen in Vermögenspreisen: Risikomodellierung, Blasenerkennung, und die Rolle der Geldpolitik
Throughout history, bursting asset price bubbles have frequently challenged not only the stability of the financial system, but have also caused severe economic contractions. Most recently, the global financial crisis (GFC) resulting from the burst of the U.S. housing bubble has provided a forceful reminder about the risks inherent in financial markets and has challenged the understanding of macro- and financial economists about the linkages between the financial system and the real economy. As a result, the crisis has also sparked intense debates about pre-crisis economic and financial policies, in particular with regard to financial market regulation and the role of monetary policy in amplifying or dampening asset price cycles. This dissertation consists of four chapters that empirically address some of these challenges and debates. Thereby, this thesis contributes to the literature on risk modeling of serially dependent asset returns; the real-time detection of asset price bubbles; forecasting of real economic activity using real-time indicators for asset price bubbles; and the role of monetary policy in asset mispricing. The first chapter, based on a paper with Helmut Herwartz and Moritz Seidel, explores whether model residuals from the class of (threshold) generalized autoregressive conditional heteroskedasticity ((T)GARCH) models are characterized by serial dependence, which could potentially be used to enhance conventional risk forecasts. We find that these residuals are hardly independent and identically distributed but instead show forms of higher order serial dependence. This suggests that TGARCH models commonly employed for predicting market risk of speculative asset returns do not use all available information for their forecast. We propose two strategies to quantify the serial dependence structures between model innovations, a nonparametric estimation approach and a flexible modeling approach based on standardized copula distributions. We show that these strategies more accurately describe the ...