This project studies new theoretical approaches for understanding the joint dynamic features of financial markets and develops econometric methods for their empirical study. In particular, it first investigates the role of heterogeneous beliefs and non-standard preferences such as prospect theory, in order to develop empirical models able to better describe the dynamic properties of credit, option, bond and equity markets. In this setting, the pricing of longevity risk and the resulting optimal risk sharing problems are also addressed. Secondly, the project proposes and validates a new general class of multivariate stochastic processes. This setup allows model builders to incorporate, in a flexible and tractable way, stochastic correlations, together with discontinuities of prices and second moments, into multivariate asset pricing models. Factor models based on this approach enable the study of large cross-sections of prices; these arise, e.g., in the risk prediction of large credit portfolios. Finally, different econometric procedures for the empirical analysis of asset pricing models are investigated. In this area, the project studies (i) nonparametric and information based methods for the estimation of general stochastic discount factor and option pricing models, (ii) robust statistical procedures with more accurate finite sample features and their application to performance evaluation and forecasting, (iii) machine learning techniques combined with boosting and bootstrap aggregation for delivering better estimates and out-of-sample predictions of the term structure of interest rates.