This thesis is composed of three chapters. The first two chapters provides novel approaches for modeling and estimating the dependence structure for a large portfolio of assets using rating data. In both chapters a natural form of organizing a portfolio in terms of the levels of exposure to economic sectors and geographical regions, plays a key role in setting up the dependence structure. The last chapter investigates weather financial strategies that exploit sector or geographical heterogeneity in the asset space are relevant in terms of portfolio optimization. This is also done in a context of a large portfolio but with data on stock returns.