This paper introduces a model-free decomposition of S&P 500 forward market index returns in terms of realized and implied dispersion,
downside, and tail risk. The decomposition utilizes a novel conditional frequency analysis on the basis of available options rather than the
times series of the S&P 500. Empirically, downside risk accounts for most of the forward market return, while symmetric tail risk is not
prominently featured. The predictable, persistent part of the realized return is very small. Nevertheless, signals revealed by this risk
anatomy provide predictive out-of-sample power for realized returns in particular for longer maturities. Furthermore, it indicates that models
with identically and independently distributed state variables are generally misspecifed in this market, and that care must be taken
when calibrating disaster risk models.