Understanding the dynamics and economic drivers of volatility, or financial assets' price variability, is of great interest to academics and practitioners. The recent literature focuses on the variance risk premium that investors require for the well known fact that volatility is stochastic. The variation over time of the magnitude of price movements of financial assets (variance risk) represents a source of uncertainty that agents are subject to. Consistently, risk adverse agents should require a compensation for bearing the randomness of future variance. The VRP is defined as the difference between the risk neutral and physical expectations of an asset's total return variation. This project aims at formulating a class of dynamic model for the latent VRP in a joint specification for the physical variance and its option implied risk neutral expectation. We advocate the inclusion of interactions and discontinuities as being essential to replicate dynamics and interdependencies between the asset's variance and its risk neutral expectation. We aim at formulating a joint specification which is computationally feasible, does not impose a restrictive parametric structure, allows to account for error in the ex-post measurement of latent variables, accounts for interactions, non-linearities and discontinuities, and linkages with macro-finance and business-cycle variables. By exploiting the temporal causality between realizations and expectations we investigate the extent to which agents anticipate large shocks on the markets (disasters) and the way they perceive such disasters in terms of the length of their impact on the markets. We also aim at extending the analysis along two dimensions: along a pure cross-sectional dimension, to various classes of financial assets (infividual stocks, diversified porfolios and commodities) in serach of idiosyncracies/commonalities and, along the expectations horizon dimension, investigating the VRP term structure.
