Asymmetric dependence modeling and implications for international diversification and risk management

We address in this dissertation the issue of asymmetric and nonlinear dependence modeling with its implications for international portfolio choice and risk management. We speak of asymmetric dependence when downside market returns are more dependent than upside market returns.;In the first chapter, we examine the problems associated with modeling the stylized fact that asset returns are more dependent in bear markets than in bull markets, called asymmetric dependence. First, we analytically show that a multivariate GARCH or regime switching model with Gaussian innovations cannot reproduce extreme dependence. We propose an alternative model which allows tail dependence for lower returns and keeps tail independence for upper returns. This model is applied to international equity and bond markets to investigate their dependence structure. It includes one normal regime in which dependence is symmetric and a second regime characterized by asymmetric dependence. Empirical results suggest that the dependence between international markets in the same class of assets is large in both regimes, while equity and bond markets exhibit little dependence, even in the same country. Exchange rate volatility appears to be related to asymmetric extreme dependence. We also use this model to analyze the lack of international portfolio diversification known as the home bias puzzle. The previous explanations for this phenomenon are relied on the first two moments: transaction costs affect the expected return, while exchange rate risk and correlation between international assets affect the volatility. Ang and Bekaert (2002) with a regime changing correlation investigate the effect of asymmetric correlation on international diversification and conclude that the costs of ignoring regimes are small. We propose an explanation based on the third moment and using a stochastic dominance argument, we prove its link with dependence asymmetry. Using the same framework, we show that asymmetric dependence amplifies the investment in the bonds, while reducing the investment in equities. This is another diversification phenomenon known as "flight to safety".;We analyze in the second chapter the implications of the asymmetric dependence on the management of extreme risks. We show that in the presence of asymmetric dependence, a portfolio model based on a multivariate symmetric GARCH with Gaussian or Student-t innovations will lead to an underestimation of the portfolio value at risk (VaR) or expected shortfall. The latter will increase when negative returns become more dependent and positive returns less dependent, while the marginal distributions are left unchanged. In fact, we show that the strong dependence for low returns increases the downside risk and this additional risk cannot be captured by the Gaussian distribution. By introducing lower tail dependence, the Student-t distribution corrects this shortcoming of the Gaussian distribution. However, the symmetric property of the Student-t means also the same dependence in the upper tail and this will reduce the downside risk. The risk model that takes into account asymmetric dependence should allow lower tail dependence and upper tail independence as put forward by Longin and Solnik (2001). The Gumbel copula captures this asymmetry and shows superiority compared to Gaussian and student-t while combined with DCC in terms of accuracy of extreme risk measures.;The third chapter addresses the problem of credibility of VaR as a risk measure in a practical context. As stressed by Artzner et al. (1999), VaR may not possess the subadditivity property required to be a coherent measure of risk. The key idea of this chapter is that, when tail thickness is responsible for violation of subadditivity, eliciting proper conditioning information may restore a VaR rationale for decentralized risk management. The argument is threefold. First, since individual traders are hired because they possess richer information on their specific market segment than senior management, they just have to follow consistently the prudential targets set by senior management to ensure that decentralized VaR control will work in a coherent way. Second, in this decentralization context, we show that if senior management has access ex-post to the portfolio shares of the individual traders, it amounts to recovering some of their private information. These shares can be used to improve backtesting in order to check that the prudential targets have been enforced by the traders. Finally, we stress that tail thickness required to violate subadditivity, even for small probabilities, remains an extreme situation since it corresponds to such poor conditioning information that expected loss appears to be infinite We then conclude that lack of coherency of decentralized VaR management, that is VaR non-subadditivity at the richest level of information, should be an exception rather than a rule. (Abstract shortened by UMI.)...