What can you Really Tell from Option Prices?

Option-implied risk-neutral moments are widely used in empirical finance to measure moment risk premia, derive bounds on expected returns, and serve as inputs for models of conditional physical moments. Their identification, however, implicitly relies on the assumption of complete option markets. We develop novel model-free bounds that characterize the valuation uncertainty that an observed option crosssection leaves about a given risk-neutral moment under market incompleteness. For the VIX, as well as for essentially all standard measures of variance, skewness, and kurtosis, this uncertainty is unbounded. Consequently, predictive regressions and cyclicality tests based on these quantities can generate almost arbitrary parameter estimates, rendering their empirical conclusions largely uninformative. As a remedy, we introduce a family of robust risk-neutral moments whose valuation uncertainty remains bounded and empirically small, while integrating into existing empirical frameworks with only minimal modification.