Fei, Fei (2013). Essays on Quantitative Risk Management. (Unpublished Doctoral thesis, City University London)
Download (1MB) | Preview
The costly lessons from global crisis in the past decade reinforce the importance as well as challenges of risk management. This thesis explores several core concepts of quantitative risk management and provides further insight.
We start with rating migration risk and propose a Mixture of Markov Chains (MMC) model to account for stochastic business cycle effects in credit rating migration risk. The model shows superior in-sample estimation and out-of-sample predication than its rivals. Compared with the naive approach the economic application suggests banks with MMC estimator will increase capital requirement in economic expansion and free up capital during recession hence it is aligned with Basel III macroprudential imitative by reducing the recession-vs-expansion gap in capital buffers.
Subsequently we move to the key concept of dependence by investigating the importance of dynamic linkages between credit and equity markets. We propose a flexible regime-switching copula model to explore the dynamics of dependence and possible structure breaks with special consideration on tail dependence. The study reveals a high-dependence regime that coincides with the recent financial crisis. The backtesting results acknowledge the new model's superiority on out-of-sample VaR forecasting over purely dynamic or static copula. It can serve to emphasise the relevance for risk management of appropriately modeling complex dependence structures.
Finally we discuss the risk measures and how they affect the portfolio optimisation. We contend that more successful portfolio management can be achieved by combining extreme value analysis to describe downside tail risk and dynamic copulas to model nonlinear dependence structures. Conditional Value-at-Risk is adopted as pertinent measure of downside tail risk for portfolio optimisation. Using both realised portfolio returns and a set of out-of-sample Monte Carlo experiments, our novel portfolio strategy is confronted with the de facto mean-variance approach. The results suggest that the MV approach produces suboptimal portfolios or a less desirable risk-return tradeoff.
|Item Type:||Thesis (Doctoral)|
|Subjects:||H Social Sciences > HG Finance|
|Divisions:||City University London PhD theses
Cass Business School > Faculty of Finance
Actions (login required)
Downloads per month over past year