Dr. T.D. (Thijs) Markwat

The dependence between large stock returns is higher than the dependence between small to moderate stock returns. This is defined as extreme dependence, and it is particularly observed for large negative returns. Therefore, diversification gains calculated from the overall dependence will overestimate the true potential for diversification during turmoil periods. This thesis answers questions on how the dependence between large negative stock returns can appropriately be modelled. The main conclusions of this thesis read that extreme dependence is often present, can become rather strong, should not be ignored, and shows substantial time-variation. More specifically, extreme dependence shows up as contagion, with small local crashes evolving into more severe crashes. In addition, due to financial globalization, and emerging market liberalization in particular, extreme dependence between regional stock markets has substantially increased. Furthermore, extreme dependence can vary over time by becoming weaker or stronger, but it can also be subject to structural changes, such as a change from symmetric dependence to asymmetric dependence. Using return data at the highest possible level of detail, improves the accuracy of forecasting joint extreme negative returns. Finally, this thesis shows how different econometric techniques can be used for modelling extreme dependence. The use of copulas for financial data is relatively new, therefore a substantial part of this thesis is devoted to new copula models and applications. Other techniques used in this thesis are GARCH, regime-switching, and logit models.