This paper o?ers a unified explanation for a host of puzzles about stocks, bonds and exchange rates. It builds on the Barro-Rietz view, that risk premia come from the probability of macroeconomic crises or disasters, and adds a variable intensity of disaster that can be asset-specific. Agents have stochastic assessments (which can be rational or behavioral) about the fundamental value that their assets would have if a disaster occurred. The model appears to explain the following puzzles on stocks, bonds and exchange rates: (i) equity premium puzzle (ii) risk-free rate-puzzle (iii) excess volatility puzzle (the fact that equity prices are so volatile) (iv) value-growth puzzle ( stocks with high price-dividend ratios have abnormally low future future) (v) upward sloping nominal yield curve (vi) Fama-Bliss findings that a higher slope of the yield curves predicts higher risk premia on bond returns (vii) corporate bond spread puzzle (the spread between corporate and government bond rates are higher than warranted by the U.S. historical experience) (viii) counter cyclical equity premium (ix) characteristics vs covariance puzzles (simple numbers such as the price-dividend ratio of stocks predict future returns better that covariances with economic factors (x) partial predictability of aggregate stock market returns by price/dividend ratios, and the consumption-asset ratio has explanatory power for future returns (xi) high price of deep out-of the money puts, and, last but not least, (xii) forward exchange rate premium puzzle (a.k.a. uncovered interest rate parity puzzle). Using the recent technique of linearity-generating processes (Gabaix 2007), the model is very tractable, and all prices are in closed form. Finally, the paper proposes a specification that integrate rare disasters with a standard real business cycle economy. In that specification, rare disasters change the asset pricing implications of the model, while keeping its business cycle implications completely unchanged.