Option Implied Dependence

We propose a novel model-free approach to infer a joint risk-neutral dependence among several assets. The dependence can be estimated when traded options are available on individ-ual assets as well as on their index. In the empirical application, we implement our approach using options on the S&P 500 index and its nine sectors. We find that option-implied dependence is highly non-normal and time-varying. Using the estimated dependence, we then study the correlation risk conditional on the market going down or up. We find that the risk premium for the down correlation is strongly negative, whereas it is positive for the up correlation. These findings are consistent with the economic intuition that the investors are particularly concerned with the loss of diversification when financial markets fall. As a result, they are willing to pay a considerable premium to hedge against increases in correlation during turbulent times. However, the investors actually prefer high correlation when markets rally.